Scene-adaptive pasture equipment centralized control method and system

Abstract

The invention discloses a scene self-adaptive pasture equipment centralized control method and system, and relates to the related field of pasture equipment intelligent control, and the method comprises the steps: carrying out the operation scene recognition according to a multi-source perception data set of a target pasture; performing adaptive classification on a pasture equipment set of the target pasture according to the N operation scene feature data; performing risk prediction on the N current control strategies of the N pasture equipment layers according to the N operation scene feature data; performing risk suppression adjustment on the N current control strategies according to the N risk prediction sequences; according to the scene control coordination degree evaluation model, performing reproduction optimization on the N scene control adjustment spaces for multiple times; conflict optimization is carried out according to the N scene control optimization strategies, and centralized control is carried out on the N pasture equipment layers according to the global control optimization strategy. The technical problem that existing pasture equipment centralized control is insufficient in control flexibility and effectiveness is solved, and the technical effect of improving the control flexibility and effectiveness is achieved.

Description

Technical Field

[0001] This application relates to the field of intelligent control of ranch equipment, and in particular to a scenario-adaptive centralized control method and system for ranch equipment. Background Technology

[0002] In modern ranch management, the ability to achieve scenario-adaptive centralized equipment control is crucial for improving ranch operational efficiency, ensuring ranch operational safety, and reducing maintenance costs. Currently, the main approach to address this issue is to employ equipment control strategies based on preset rules or simple conditional judgments. These strategies attempt to meet the ranch's operational needs by controlling ranch equipment in single or limited scenarios. However, these current methods lack deep perception and intelligent analysis capabilities for complex and ever-changing ranch operational scenarios. Consequently, the control strategies cannot be flexibly adjusted to different operational scenarios, making it difficult to effectively address various potential risks. This results in poor equipment control performance and may even lead to safety accidents.

[0003] At present, the centralized control of ranch equipment suffers from technical problems such as insufficient control flexibility and effectiveness. Summary of the Invention

[0004] This application provides a scenario-adaptive centralized control method and system for ranch equipment. It identifies operational scenarios using a multi-source sensing dataset, obtaining N scenario feature data. Based on these feature data, the ranch equipment set is adaptively classified to form N corresponding ranch equipment layers. Risk prediction is performed on the current control strategy of each equipment layer based on the operational scenario feature data, generating N risk prediction sequences. These sequences are used to adjust the control strategy for risk suppression, constructing N scenario control adjustment spaces. A scenario control coordination evaluation model is used to perform multiple iterations of optimization within the adjustment spaces, generating N scenario control optimization strategies. These optimization strategies are then conflict-optimized and integrated to generate a global control optimization strategy. Based on this, centralized control is implemented for all ranch equipment layers. These technical means solve the technical problems of insufficient control flexibility and effectiveness in existing centralized control of ranch equipment, achieving the technical effect of improving control flexibility and effectiveness.

[0005] This application provides a scenario-adaptive centralized control method for ranch equipment, comprising: identifying operational scenarios based on a multi-source sensing dataset of a target ranch to obtain N operational scenario feature data, where N is a positive integer greater than 1; adaptively classifying the ranch equipment set of the target ranch based on the N operational scenario feature data to obtain N ranch equipment layers; predicting the risks of N current control strategies of the N ranch equipment layers based on the N operational scenario feature data to obtain N risk prediction sequences; adjusting the risk suppression of the N current control strategies based on the N risk prediction sequences to generate N scenario control adjustment spaces; performing multiple breeding optimizations on the N scenario control adjustment spaces based on a scenario control coordination degree evaluation model to generate N scenario control optimization strategies; performing conflict optimization based on the N scenario control optimization strategies to obtain a global control optimization strategy; and centrally controlling the N ranch equipment layers based on the global control optimization strategy.

[0006] In a possible implementation, risk prediction is performed on the N current control strategies of the N pasture equipment layers based on the N operational scenario feature data to obtain N risk prediction sequences. The following processing is then performed: Based on the N operational scenario feature data and the N current control strategies, the nth operational scenario feature data and the nth current control strategy corresponding to the nth pasture equipment layer are extracted, where n is a positive integer, 1≤n≤N; Based on the nth operational scenario feature data, scenario control mismatch risk prediction is performed on the nth current control strategy to obtain the nth scenario control mismatch risk coefficient; Based on the nth operational scenario feature data, livestock induction risk prediction is performed on the nth current control strategy to obtain the nth induction livestock risk coefficient; Based on the nth operational scenario feature data, equipment failure risk prediction is performed on the nth current control strategy to obtain the nth equipment failure risk coefficient; Combining the nth scenario control mismatch risk coefficient and the nth induction livestock risk coefficient, the nth risk prediction sequence is generated.

[0007] In a possible implementation, based on the feature data of the nth operation scenario, the scenario control mismatch risk is predicted for the nth current control strategy to obtain the nth scenario control mismatch risk coefficient. The following processing is then performed: a scenario control mismatch risk record retrieval is performed on the target ranch to obtain a control strategy sample set and a scenario control mismatch risk sample set; the scenario control mismatch risk sample set is clustered according to the control strategy sample set to obtain each scenario control mismatch risk group corresponding to each control strategy sample; confidence fusion is performed on each scenario control mismatch risk group to obtain a mismatch risk sample confidence set; a risk prediction learner is supervised and trained according to the control strategy sample set and the mismatch risk sample confidence set, and a mismatch risk prediction loss coefficient is obtained after each predetermined number of training iterations; if the mismatch risk prediction loss coefficient is less than the mismatch risk prediction loss threshold, a scenario control mismatch risk prediction model is obtained; the feature data of the nth operation scenario and the nth current control strategy are input into the scenario control mismatch risk prediction model to obtain the nth scenario control mismatch risk coefficient.

[0008] In a possible implementation, risk suppression and adjustment are performed on the N current control strategies based on the N risk prediction sequences to generate N scenario control adjustment spaces. The following processes are then performed: It is determined whether the nth risk prediction sequence satisfies multivariate risk constraints, including scenario control mismatch risk constraints, livestock induced risk constraints, and equipment failure risk constraints; if the nth risk prediction sequence does not satisfy the multivariate risk constraints, control scheme retrieval is performed on the nth pasture equipment layer based on the nth operation scenario feature data to establish an nth control scheme retrieval set; multidimensional control trigger feature analysis is performed on the nth pasture equipment layer based on the nth control scheme retrieval set to obtain the nth control trigger domain; multidimensional adjustment is performed on the nth current control strategy based on the nth control trigger domain to obtain the nth control adjustment group; multidimensional risk joint optimization is performed on the nth control adjustment group based on the multivariate risk constraints to generate the nth scenario control adjustment space.

[0009] In a possible implementation, multi-dimensional risk joint optimization is performed on the nth control and regulation group based on the multi-dimensional risk constraints to generate an nth scenario control and regulation space, and the following processing is performed: Based on the nth control and regulation group, a first control and regulation scheme is extracted; based on the feature data of the nth operation scenario, multi-dimensional risk prediction is performed on the first control and regulation scheme to obtain a first scheme risk prediction sequence; it is determined whether the first scheme risk prediction sequence satisfies the multi-dimensional risk constraints; if the first scheme risk prediction sequence satisfies the multi-dimensional risk constraints, the first control and regulation scheme is added to the nth scenario control and regulation space.

[0010] In a possible implementation, the N scene control adjustment spaces are subjected to multiple breeding optimizations based on the scene control coordination degree evaluation model to generate N scene control optimization strategies. The following processes are then performed: Based on the feature data of the nth job scene, the scene control coordination degree of the nth scene control adjustment space is analyzed according to the scene control coordination degree evaluation model to obtain a scene control coordination degree evaluation set; based on the scene control coordination degree evaluation set, the nth scene control adjustment space is optimized and screened according to a predetermined scene control coordination degree to establish an initial optimization group for the nth scene control; the initial optimization group for the nth scene control is bred and optimized according to the scene control coordination degree evaluation model and the predetermined scene control coordination degree to obtain a first bred optimization group for scene control; the first bred optimization group for scene control is further bred and optimized according to the scene control coordination degree evaluation model and the predetermined scene control coordination degree to obtain a Qth bred optimization group for scene control, where Q is a positive integer greater than 1; energy consumption minimization optimization is performed based on the initial optimization group for the nth scene control, the first bred optimization group for scene control, ..., the Qth bred optimization group for scene control to generate an nth scene control optimization strategy.

[0011] In a possible implementation, the initial optimization group of the nth scene control is subjected to breeding optimization based on the scene control coordination degree evaluation model and the predetermined scene control coordination degree to obtain a first breeding optimization group of scene control. The following processes are then performed: the breeding capacity of the initial optimization group of the nth scene control is allocated according to the breeding capacity constraint to obtain the breeding capacity allocation result; based on the breeding capacity allocation result, the initial optimization group of the nth scene control is bred according to the nth control trigger domain to obtain a first breeding group of scene control; the first breeding group of scene control is subjected to multi-dimensional risk joint optimization based on multi-dimensional risk constraints to obtain a first breeding optimization group of scene control; and the first breeding optimization group of scene control is generated by optimizing the scene control coordination degree based on the scene control coordination degree evaluation model and the predetermined scene control coordination degree of the first breeding optimization group of scene control.

[0012] In a possible implementation, the operation scenario is identified based on the multi-source sensing dataset of the target pasture, and the following processing is performed: multi-source monitoring of the target pasture is performed to obtain a pasture monitoring dataset; the pasture monitoring dataset is cleaned to generate the multi-source sensing dataset.

[0013] In a possible implementation, N ranch equipment layers are obtained, and the following processing is performed: real-time control parameters of each device in the ranch equipment set are collected to obtain an equipment control dataset; the equipment control dataset is clustered according to the N ranch equipment layers to generate the N current control strategies.

[0014] This application also provides a scenario-adaptive centralized control system for ranch equipment, comprising: a scenario identification module for identifying scenarios based on a multi-source sensing dataset of a target ranch, obtaining N scenario feature data, where N is a positive integer greater than 1; an equipment adaptive classification module for adaptively classifying the ranch equipment set of the target ranch based on the N scenario feature data, obtaining N ranch equipment layers; a control strategy risk prediction module for predicting the risks of N current control strategies of the N ranch equipment layers based on the N scenario feature data, obtaining N risk prediction sequences; a risk suppression and adjustment module for performing risk suppression and adjustment on the N current control strategies based on the N risk prediction sequences, generating N scenario control adjustment spaces; an optimization module for performing multiple breeding optimizations on the N scenario control adjustment spaces based on a scenario control coordination degree evaluation model, generating N scenario control optimization strategies; and a conflict optimization module for performing conflict optimization based on the N scenario control optimization strategies, obtaining a global control optimization strategy, and performing centralized control on the N ranch equipment layers based on the global control optimization strategy.

[0015] This application proposes a scenario-adaptive centralized control method and system for ranch equipment. First, it identifies operational scenarios based on a multi-source sensing dataset of the target ranch, obtaining N operational scenario feature data (N being a positive integer greater than 1). Next, it adaptively classifies the ranch equipment set based on these N operational scenario feature data, obtaining N ranch equipment layers. Then, it predicts the risks of N current control strategies for the N ranch equipment layers based on the N operational scenario feature data, obtaining N risk prediction sequences. Following this, it adjusts the risk suppression of the N current control strategies based on the N risk prediction sequences, generating N scenario control adjustment spaces. Then, it performs multiple iterations of optimization on these N scenario control adjustment spaces using a scenario control coordination evaluation model, generating N scenario control optimization strategies. Finally, it performs conflict optimization based on these N scenario control optimization strategies to obtain a global control optimization strategy, and centrally controls the N ranch equipment layers based on this global control optimization strategy. This achieves the technical effect of improving control flexibility and effectiveness. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a flowchart illustrating a scenario-adaptive centralized control method for ranch equipment provided in an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of a scenario-adaptive centralized control system for ranch equipment provided in an embodiment of this application.

[0019] Figure labeling: 10 for work scene recognition module, 20 for equipment adaptive classification module, 30 for control strategy risk prediction module, 40 for risk suppression and adjustment module, 50 for optimization module, and 60 for conflict optimization module. Detailed Implementation

[0020] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below.

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, 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 explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0023] This application provides a scenario-adaptive centralized control method for ranch equipment, such as... Figure 1 As shown, the method includes: Step S100: Based on the multi-source perception dataset of the target ranch, identify the operation scene and obtain N operation scene feature data, where N is a positive integer greater than 1.

[0024] Specifically, the multi-source sensing dataset refers to a collection of various data related to ranch operation scenarios collected through multiple types of sensors, including information on the environment, animals, equipment, and other aspects. The operation scenario feature data refers to key data features extracted from the multi-source sensing dataset that characterize specific ranch operation scenarios, used to distinguish different operation scenarios and subsequent targeted equipment control.

[0025] Multi-source data acquisition technology is employed, utilizing various sensors installed on the target ranch, such as light sensors, temperature and humidity sensors, soil moisture sensors, and video cameras, to collect multi-source sensing data related to the ranch operation scenario. This includes environmental data (light intensity, humidity, temperature, wind speed, soil moisture, etc.), animal data (location, activity level, feeding status, etc. of cattle and sheep), and equipment data (equipment operating status, working duration, etc.). Then, data fusion technology and machine learning algorithms, such as classification algorithms like Support Vector Machines (SVM) and decision trees, are used to identify the operation scenario from the fused data, extracting N operation scenario feature data. These feature data could represent grazing scenarios where light intensity is within a certain range and humidity is within a specific interval, or irrigation scenarios where temperature is low and soil moisture is high, etc.

[0026] For example, the target ranch is a large cattle ranch with different functional areas such as grazing areas, feeding areas, barns, and watering areas. It is equipped with numerous devices including fence gate control equipment, automatic feeding equipment, environmental control equipment (fans, water curtains, etc.), and water pumps. Light sensors and video cameras are installed in the grazing area to collect data on light intensity and cattle activity. Weight sensors are installed in the feeding area to monitor the amount of remaining hay, while cameras observe the cattle's feeding behavior. Temperature and humidity sensors are placed in the barn to acquire environmental temperature and humidity data. After fusing this data, a convolutional neural network (CNN) algorithm from deep learning is used to identify the characteristic data corresponding to N operational scenarios, such as grazing, feeding, and barn environmental control. For example, the characteristics of the grazing scenario include high light intensity, cattle activity in the grazing area, and an increase in the weight of hay in the feeding area.

[0027] In one possible implementation, operational scenario identification is performed based on a multi-source sensing dataset of the target pasture. Step S100 further includes step S110, which involves multi-source monitoring of the target pasture to obtain a pasture monitoring dataset. Specifically, various sensors are deployed on the target pasture, such as weather stations collecting meteorological data (temperature, humidity, wind speed, rainfall, etc.), soil sensors monitoring soil moisture and nutrients, video surveillance equipment acquiring real-time images of various areas of the pasture, positioning and physiological monitoring sensors worn by animals such as cattle and sheep collecting information such as animal location, activity level, and body temperature, as well as the equipment's own sensors monitoring equipment operating status (motor speed, equipment load, operating time, etc.). This allows for comprehensive and multi-dimensional real-time monitoring of the pasture, thereby obtaining a pasture monitoring dataset containing various types of information. These data come from diverse sources and are rich in type, including various aspects such as the pasture environment, animal activity, and equipment operation, providing a comprehensive data foundation for subsequent operational scenario identification.

[0028] Step S120: Data cleaning is performed on the pasture monitoring dataset to generate the multi-source sensing dataset. Specifically, the original data in the pasture monitoring dataset is cleaned. First, duplicate data is removed. For example, the same data packets repeatedly sent by the same sensor within a short period of time may be due to unstable sensor signals or communication failures. By setting the tolerance range for timestamps and data values, duplicate records are identified and deleted to reduce the impact of redundant data on subsequent processing and improve data processing efficiency. Second, missing data is processed. For data missing due to sensor failures or network transmission interruptions, interpolation methods are used to fill in the missing data. For example, for missing temperature values ​​in meteorological data, the temperature value at the missing time can be estimated using linear interpolation based on temperature data at adjacent time points before and after the monitoring point. For missing location data in animal activity trajectories, a motion prediction model can be used to supplement the data based on animal movement patterns and information from adjacent location points, ensuring the integrity and continuity of the data sequence and avoiding significant deviations in the identification of operational scenarios due to missing data.

[0029] This approach effectively removes duplicate data and processes missing data through data cleaning, reducing the interference of data noise and incompleteness on the identification results. This allows the operational scenario feature data to more realistically and accurately reflect the actual operational scenarios of the ranch, thereby improving the accuracy of operational scenario identification. This lays the foundation for subsequent steps such as equipment classification, risk prediction, and control strategy optimization for different operational scenarios, ensuring that the entire ranch's centralized equipment control system can make precise decisions and controls based on accurate scenario information.

[0030] Step S200: Adaptively classify the ranch equipment set of the target ranch according to the N operation scenario feature data to obtain N ranch equipment layers.

[0031] Specifically, the ranch equipment layer is a hierarchical structure formed by classifying ranch equipment according to the characteristic data of the operation scenario. Each equipment layer contains a set of equipment related to a specific operation scenario, which is used to formulate targeted control strategies.

[0032] Based on the characteristic data of the operation scenarios, clustering algorithms, such as the K-Means algorithm, are used to classify the ranch equipment set according to its adaptability to different operation scenarios, dividing it into N ranch equipment layers. For example, in the grazing scenario, grazing-related equipment such as fence gate control equipment and grazing railcars are divided into one layer; in the irrigation scenario, irrigation sprinklers, water pumps, and other equipment are divided into one layer.

[0033] For example, for grazing scene feature data, the K-Means algorithm is used to classify the pasture equipment set, classifying fence gate control equipment, automatic railcars used to drive cattle herds, etc., into the grazing scene equipment layer; automatic feeding equipment into the feeding scene equipment layer; and environmental control equipment such as fans and water curtains in the cattle shed area into the cattle shed environmental control scene equipment layer.

[0034] In one possible implementation, N ranch equipment layers are obtained. Step S200 further includes step S210, which involves collecting real-time control parameters of each device within the ranch equipment set to obtain an equipment control dataset. Specifically, real-time control parameters of each device within the ranch equipment set are collected using remote monitoring systems, the devices' own sensors, and on-site data acquisition terminals to obtain the equipment control dataset. These real-time control parameters include the device's operating power, speed, operating time, on / off status, and operating mode. For example, for automatic feeding equipment, parameters such as feeding frequency, single feeding amount, and operating time are collected; for water pump equipment, parameters such as flow rate, pressure, and start / stop time are collected, comprehensively reflecting the current operating status and control of the equipment, providing detailed data support for subsequent cluster analysis.

[0035] Step S220: Cluster the equipment control dataset according to the N pasture equipment layers to generate the N current control strategies. Specifically, the equipment control dataset is clustered based on the pre-identified N pasture equipment layers. Specifically, first, key features in each equipment control dataset are determined, such as the aforementioned operating power and rotational speed. Then, a suitable clustering algorithm, such as the K-Means algorithm or hierarchical clustering algorithm, is selected to divide the equipment control dataset into N categories corresponding to the N pasture equipment layers based on these key features. Each clustered category represents a current control strategy, which includes a combination of control parameters applicable to the corresponding pasture equipment layer. For example, for equipment layers in a grazing scenario, the clustered control strategy may include parameters such as the opening and closing interval of the fence gate and the operating speed of the grazing track vehicle. These parameter combinations have proven to be suitable for the control method of this equipment layer in the corresponding scenario in past operations.

[0036] This approach, by collecting real-time control parameters from each device, enables a precise understanding of the actual operating status and control requirements of each device. Based on this, clustering is performed according to the ranch equipment layers to generate the current control strategy. This fully considers the characteristics and differences of different equipment layers, ensuring that each control strategy closely matches the operating rules and requirements of the corresponding equipment layer in a specific operational scenario, thereby improving the relevance and adaptability of the control strategy.

[0037] Step S300: Based on the N operational scenario feature data, perform risk prediction on the N current control strategies of the N ranch equipment layers to obtain N risk prediction sequences.

[0038] Specifically, the risk prediction sequence is a sequence of data obtained by predicting the probability of risks occurring in the current control strategy of the equipment in the future based on a risk prediction model, and is used to guide the adjustment of the control strategy.

[0039] Using time series analysis methods, such as the ARIMA model, and combining historical equipment failure data with operational scenario characteristic data, risk prediction is performed on the current control strategy for each farm's equipment layer. For example, the probability of a certain irrigation equipment failing when there were sudden changes in light intensity and prolonged equipment operation in the past is analyzed to generate a corresponding risk prediction sequence, which includes numerical values ​​indicating the likelihood of risk occurring at different time points.

[0040] For example, for the fence gate control equipment in the grazing scenario, we collect its past operating data and fault records under different lighting and cattle activity conditions. We use the ARIMA model to predict the probability of the fence gate failing due to risks such as motor overheating during subsequent operations, and generate a risk prediction sequence. For example, the risk values ​​of failure predicted in the next hour are 0.1, 0.15, and 0.2, respectively.

[0041] In one possible implementation, risk prediction is performed on the N current control strategies of the N pasture equipment layers based on the N operational scenario feature data to obtain N risk prediction sequences. Step S300 further includes step S310, which extracts the nth operational scenario feature data and the nth current control strategy corresponding to the nth pasture equipment layer based on the N operational scenario feature data and the N current control strategies, where n is a positive integer, 1≤n≤N. Specifically, the nth operational scenario feature data and the nth current control strategy corresponding to the nth pasture equipment layer are extracted from the previously obtained N operational scenario feature data and N current control strategies. For example, when the nth pasture equipment layer is an irrigation equipment layer, feature data related to the irrigation operation scenario, such as soil moisture and weather conditions, and the control strategies of the current irrigation equipment, such as irrigation time interval and irrigation water volume parameter settings, are extracted.

[0042] Step S320: Based on the feature data of the nth operation scenario, predict the scenario control mismatch risk of the nth current control strategy to obtain the nth scenario control mismatch risk coefficient. Specifically, based on the feature data of the nth operation scenario, machine learning algorithms such as logistic regression or decision tree models are used to predict the scenario control mismatch risk of the nth current control strategy. By analyzing historical data on scenario control mismatch situations caused by unreasonable control strategies under similar operation scenario characteristics (such as irrigation when soil moisture is already suitable, resulting in excessive water supply), the nth scenario control mismatch risk coefficient is calculated. This coefficient reflects the likelihood of adverse effects caused by the mismatch between the current control strategy and the operation scenario.

[0043] Step S330: Based on the feature data of the nth work scenario, predict the livestock risk induced by the nth current control strategy to obtain the nth induced livestock risk coefficient. Specifically, based on the feature data of the nth work scenario, animal behavior models and risk assessment algorithms are used to predict the livestock risk induced by the nth current control strategy. For example, in a grazing scenario, based on the control strategy of the fence gate opening and closing (such as opening and closing speed, interval time, etc.) and the activity characteristics and location information of cattle and sheep, predict the probability of livestock being startled or injured by colliding with the fence due to improper control, and obtain the nth induced livestock risk coefficient.

[0044] Step S340: Based on the feature data of the nth operation scenario, predict the equipment failure risk of the nth current control strategy to obtain the nth equipment failure risk coefficient. Combine the control mismatch risk coefficient of the nth scenario and the induced livestock risk coefficient to generate the nth risk prediction sequence. Specifically, also based on the feature data of the nth operation scenario, combined with the equipment's operating status data and fault history records, a reliability analysis method and a predictive maintenance model are used to predict the equipment failure risk under the nth current control strategy. For example, for automatic feeding equipment, considering its operating frequency, load, and environmental humidity in the feeding operation scenario, predict the probability of motor failure, conveyor belt jamming, and other failures to obtain the nth equipment failure risk coefficient. Finally, by combining the risk coefficients of the control mismatch in the nth scenario, the risk coefficient of the livestock induced by the nth scenario, and the risk coefficient of the equipment failure in the nth scenario, a risk prediction sequence is generated according to certain combination rules. This sequence reflects the changing trends of the risk of control mismatch in the nth pasture equipment layer under the current control strategy at different time points or under different conditions.

[0045] This approach, through in-depth analysis of operational scenario characteristic data and current control strategies, predicts risks from three dimensions: scenario control mismatch, induced livestock risks, and equipment failure risks. It can comprehensively and accurately identify various risk factors that ranch equipment may face during operation, providing an accurate basis for subsequent targeted risk prevention and control measures, thereby effectively improving the safety and stability of ranch equipment operation.

[0046] In one possible implementation, based on the feature data of the nth operation scenario, the current control strategy for the nth operation scenario is predicted for scenario control mismatch risk to obtain the nth scenario control mismatch risk coefficient. Step S320 further includes step S321, which involves retrieving scenario control mismatch risk records for the target pasture to obtain a control strategy sample set and a scenario control mismatch risk sample set. Specifically, historical data of the target pasture is retrieved to find records related to scenario control mismatch risk. These records include the control strategies adopted by the pasture equipment under different operation scenarios and the corresponding scenario control mismatch risk. For example, records show that under specific soil moisture, temperature, and other operation scenario characteristics, when using a certain irrigation equipment control strategy, scenario control mismatch risks such as uneven irrigation occurred, thereby obtaining a sample set containing multiple control strategy samples and corresponding multiple historical scenario control mismatch risk coefficients. Each control strategy sample may correspond to multiple historical scenario control mismatch risk coefficients because the mismatch risk of the same control strategy may be different under different historical scenarios.

[0047] Step S322: Cluster the scenario control mismatch risk sample set according to the control strategy sample set to obtain each scenario control mismatch risk group corresponding to each control strategy sample. Specifically, the control strategy sample set and the scenario control mismatch risk sample set are integrated together, and the scenario control mismatch risk sample set is clustered based on the control strategy samples. For example, the K-Means clustering algorithm is used to group similar control strategy samples and their corresponding scenario control mismatch risk coefficients into one category according to the characteristics of the control strategy samples (such as irrigation time interval, irrigation water volume, etc.), forming multiple scenario control mismatch risk groups. The control strategy samples in each risk group have similar characteristics, and the scenario control mismatch risk coefficients corresponding to these samples also have a certain similarity.

[0048] Step S323 involves performing confidence fusion on the control mismatch risk groups for each scenario to obtain a confidence set of mismatch risk samples. Specifically, confidence fusion is performed separately for each scenario control mismatch risk group. Confidence fusion is a data fusion technique that comprehensively considers the credibility of control mismatch risk coefficients for various historical scenarios. Through a specific algorithm (such as weighted averaging, where weights are determined based on factors like data source reliability and data freshness), multiple risk coefficients are merged into a single confidence mismatch risk coefficient, thereby obtaining a confidence set of mismatch risk samples. This step reduces data redundancy and noise, improving data reliability and representativeness.

[0049] Step S324: Supervised training of the risk prediction learner is performed based on the control strategy sample set and the mismatch risk sample confidence set. After each predetermined number of training iterations, the mismatch risk prediction loss coefficient is obtained. Specifically, a risk prediction learner (such as a neural network, support vector machine, or other machine learning model) is constructed, and the control strategy sample set and the mismatch risk sample confidence set are input as training data into the model for supervised training. After each predetermined number of training iterations (e.g., every 100 training iterations), the difference between the model's output and the actual confidence mismatch risk coefficient is calculated to obtain the mismatch risk prediction loss coefficient. The loss coefficient reflects the magnitude of the error between the model's prediction result and the true value; the smaller the error, the better the model's prediction performance.

[0050] Step S325: If the mismatch risk prediction loss coefficient is less than the mismatch risk prediction loss threshold, a scenario control mismatch risk prediction model is obtained. Specifically, a mismatch risk prediction loss threshold is set. When the mismatch risk prediction loss coefficient is less than this threshold, it indicates that the model's prediction accuracy has met the expected requirements. At this point, training is complete, and the scenario control mismatch risk prediction model is obtained. This model can predict the scenario control mismatch risk coefficient based on the input operational scenario feature data and the current control strategy.

[0051] Step S326: Input the nth operation scenario feature data and the nth current control strategy into the scenario control mismatch risk prediction model to obtain the nth scenario control mismatch risk coefficient. Specifically, the nth operation scenario feature data and the nth current control strategy are input as input feature vectors into the scenario control mismatch risk prediction model. After internal calculation and analysis, the model outputs the nth scenario control mismatch risk coefficient, which represents the probability of scenario control mismatch risk occurring when using the nth current control strategy in the nth operation scenario.

[0052] This approach deeply mines and processes historical scenario control mismatch risk data through retrieval, clustering, and confidence fusion, removing noise and redundant information to make the training data more accurate and reliable. Simultaneously, supervised training continuously optimizes the parameters of the risk prediction learner until the model's prediction loss coefficient meets requirements. This effectively improves the accuracy of scenario control mismatch risk prediction, providing a more accurate basis for subsequent risk mitigation and control strategy optimization, and reducing control errors and potential losses caused by inaccurate risk prediction.

[0053] Step S400: Adjust the risk suppression of the N current control strategies according to the N risk prediction sequences to generate N scenario control adjustment spaces.

[0054] Specifically, the scenario control adjustment space is the range and possibility of adjusting the equipment control strategy based on the risk prediction sequence, which is used to provide operational space for subsequent optimization.

[0055] Based on the risk prediction sequence, fuzzy control technology is used to construct a fuzzy rule base, such as "if the risk prediction value is high, then increase the control adjustment range," to perform risk suppression adjustment on the current control strategy. For each farm equipment layer, a corresponding scenario control adjustment space is established to determine the range of control parameters that can be adjusted for the equipment under different risk levels, such as the adjustment range of equipment operating speed, start-up and shutdown time, etc.

[0056] For example, based on the risk prediction sequence of the fence gate, fuzzy rules are constructed. When the risk prediction value exceeds 0.15, the opening and closing frequency of the fence gate is reduced and the interval time is extended, thereby adjusting the risk suppression of the current control strategy of the device and forming a scene control adjustment space for the fence gate in the grazing scenario. That is, its opening and closing frequency can be adjusted within a certain range and the interval time can be extended accordingly.

[0057] In one possible implementation, risk suppression adjustment is performed on the N current control strategies based on the N risk prediction sequences to generate N scenario control adjustment spaces. Step S400 further includes step S410, determining whether the nth risk prediction sequence satisfies multivariate risk constraints, which include scenario control mismatch risk constraints, livestock induction risk constraints, and equipment failure risk constraints. Specifically, for the nth pasture equipment layer, the system simultaneously checks whether the nth risk prediction sequence simultaneously satisfies scenario control mismatch risk constraints, livestock induction risk constraints, and equipment failure risk constraints. These three risk constraints respectively set risk thresholds for scenario control mismatch, livestock safety, and equipment failure. For example, at the irrigation equipment layer, the scenario control mismatch risk constraint can be a risk coefficient of less than 0.2 for substandard irrigation uniformity; the livestock induction risk constraint is a risk coefficient of less than 0.1 for livestock being startled by stimuli such as water flow noise during irrigation; and the equipment failure risk constraint is a risk coefficient of less than 0.3 for equipment failures such as water pump malfunctions in the irrigation equipment.

[0058] Step S420: If the nth risk prediction sequence does not meet the multivariate risk constraints, a control scheme retrieval is performed on the nth pasture equipment layer based on the nth operation scenario characteristic data to establish an nth control scheme retrieval set. Specifically, if the nth risk prediction sequence does not meet the multivariate risk constraints, the system will perform a control scheme retrieval on the nth pasture equipment layer based on the nth operation scenario characteristic data. For example, for the fence gate equipment layer in a grazing scenario, the system will search the control scheme database for multiple control schemes that match the characteristics of the current grazing scenario (such as the number of cattle and sheep, the grazing area, wind speed, etc.). These control schemes include schemes with different combinations of parameters such as fence gate opening and closing speed and opening and closing frequency, thereby establishing an nth control scheme retrieval set.

[0059] Step S430: Perform multi-dimensional control trigger feature analysis on the nth pasture equipment layer based on the nth control scheme retrieval set to obtain the nth control trigger domain. Specifically, the system performs multi-dimensional control trigger feature analysis on each control scheme in the nth control scheme retrieval set. Multi-dimensional control trigger features refer to key factors that can trigger changes in equipment control strategies. Taking the automatic feeding equipment layer as an example, control trigger features may include time (such as feeding time interval), cattle and sheep activity status (such as aggregation degree), and remaining feed amount. The system analyzes the value range and combination of these control trigger features under different control schemes to obtain the nth control trigger domain, that is, to determine under what conditions (the value range of the trigger features) the corresponding control scheme will be activated.

[0060] Step S440: The current control strategy is multidimensionally adjusted based on the nth control trigger domain to obtain the nth control adjustment group. Specifically, the current control strategy is multidimensionally adjusted based on the nth control trigger domain. For example, in the cattle shed environmental control equipment layer, if the control trigger domain analysis finds that the fan speed and water curtain opening degree need to be adjusted when the temperature exceeds 25℃ and the humidity exceeds 60%, the system will adjust the current fan speed and water curtain opening degree control strategy under these trigger conditions, generating a series of adjusted control strategy combinations, forming the nth control adjustment group. These combinations include different fan speed and water curtain opening degree combinations to adapt to different temperature and humidity conditions.

[0061] Step S450: Based on the multivariate risk constraints, perform multidimensional risk joint optimization on the nth group of control and regulation to generate the nth scenario control and regulation space. Specifically, the system performs multidimensional risk joint optimization on the nth group of control and regulation based on multivariate risk constraints. This process comprehensively considers scenario control mismatch risk, induced livestock risk, and equipment failure risk, searching for combinations of control strategies within the nth group of control and regulation that meet the risk constraints. For example, at the milking equipment layer, the system will find multiple combinations of control strategies, thereby generating the nth scenario control and regulation space, under the constraints of scenario control mismatch risk (such as the risk of matching milking speed with cow milk production speed), induced livestock risk (risk of stress response to cows during milking), and equipment failure risk (risk of failure of milking equipment components).

[0062] This approach, through the scenario control adjustment space generated by multi-dimensional risk joint optimization, comprehensively considers various risk factors such as scenario control mismatch, inducing livestock and equipment failures. This enables the control strategy of ranch equipment to operate under multiple risk constraints, effectively reducing various risks that may occur in the production and operation of ranches.

[0063] In one possible implementation, multi-dimensional risk joint optimization is performed on the nth control and regulation group based on the multi-dimensional risk constraints to generate the nth scenario control and regulation space. Step S450 further includes step S451, extracting a first control and regulation scheme based on the nth control and regulation group. Specifically, a control and regulation scheme is extracted from the nth control and regulation group as the first control and regulation scheme. For example, when adjusting the control strategy of the automatic milking equipment layer in a pasture, the nth control and regulation group contains multiple schemes with different parameter combinations, such as schemes for adjusting milking speed, milking pressure, etc. One scheme is selected as the first scheme. This scheme can be determined based on a certain priority (such as high frequency of use in the past, low equipment energy consumption, etc.) or randomly selected.

[0064] Step S452: Based on the feature data of the nth operation scenario, perform multi-dimensional risk prediction on the first control and regulation scheme to obtain a risk prediction sequence for the first scheme. Specifically, based on the feature data of the nth operation scenario, perform multi-dimensional risk prediction on the first control and regulation scheme. The operation scenario feature data includes environmental factors (such as temperature, humidity, etc.), livestock status (such as the health status of dairy cows, milk production, etc.), and equipment operating status (such as the wear and tear of milking equipment components, etc.). Using these data and the constructed risk prediction model, predict the risk values ​​of the scheme in three dimensions: scenario control mismatch risk, induced livestock risk, and equipment failure risk, thereby obtaining a risk prediction sequence for the first scheme. For example, predicting that under the current operation scenario, when using the first control and regulation scheme, the scenario control mismatch risk coefficient is 0.15 (indicating a certain probability that the milking speed and the milk production speed of dairy cows will not match), the induced livestock risk coefficient is 0.1 (indicating a certain probability that the operation during the milking process will cause stress in dairy cows), and the equipment failure risk coefficient is 0.2 (indicating a certain probability that milking equipment components will fail).

[0065] Step S453: Determine whether the first scheme risk prediction sequence satisfies the multivariate risk constraints. Specifically, compare each risk value in the first scheme risk prediction sequence with the corresponding threshold in the multivariate risk constraints to determine whether the constraints are met. For example, the multivariate risk constraints set the scenario control mismatch risk constraint threshold to 0.2, the induced livestock risk constraint threshold to 0.15, and the equipment failure risk constraint threshold to 0.25. If the scenario control mismatch risk coefficient 0.15 in the first scheme risk prediction sequence is less than 0.2, the induced livestock risk coefficient 0.1 is less than 0.15, and the equipment failure risk coefficient 0.2 is less than 0.25, then the first scheme risk prediction sequence is determined to satisfy the multivariate risk constraints.

[0066] Step S454: If the first scheme risk prediction sequence satisfies the multivariate risk constraints, the first control and regulation scheme is added to the nth scenario control and regulation space. Specifically, if the first scheme risk prediction sequence satisfies the multivariate risk constraints, it indicates that the scheme is a feasible control strategy that can operate within a controllable risk range. Therefore, the first control and regulation scheme is added to the nth scenario control and regulation space. For example, in the feed mixing equipment layer of a ranch, the first control and regulation scheme that satisfies the multivariate risk constraints after verification through the above steps is added to the scenario control and regulation space of that equipment layer, providing a feasible equipment control strategy option for ranch management for subsequent centralized control decisions.

[0067] This implementation method ensures that only control schemes that have been verified and meet risk constraints are added to the scenario control adjustment space by performing multi-dimensional risk prediction and judgment on each control and adjustment scheme individually. This guarantees that each control strategy in the scenario control adjustment space has high reliability and security. Appropriate control strategies can be quickly selected based on this high-quality scenario control adjustment space, reducing the risks caused by inappropriate scheme selection and improving the overall performance and stability of the ranch equipment centralized control system.

[0068] Step S500: Based on the scene control coordination degree evaluation model, perform multiple breeding optimizations on the N scene control adjustment spaces to generate N scene control optimization strategies.

[0069] Specifically, the scenario control coordination evaluation model is used to evaluate the degree of cooperation and coordination between different scenario control strategies, so as to ensure that the operating strategies of devices in each scenario can cooperate with each other to achieve the best overall control effect.

[0070] This paper employs a genetic algorithm, an intelligent optimization algorithm, using scene control coordination degree as the fitness function, to perform multiple breeding optimizations on control strategies within N scene control adjustment spaces. Specifically, through genetic operations such as selection, crossover, and mutation, a new population of control strategies is continuously generated iteratively until the N optimal scene control strategies with the highest scene control coordination degree are found.

[0071] For example, guided by a scenario control coordination evaluation model, a genetic algorithm is used to optimize control strategies within the control adjustment space of each scenario. For instance, the initial population consists of individuals with control strategies based on combinations of control parameters such as different fence gate opening / closing frequencies, feeding amounts from feeding equipment, and fan speeds. After multiple generations of genetic operations, N scenario control optimization strategies with high scenario coordination are selected, such as strategies for appropriate fence gate opening / closing frequencies and intervals in grazing scenarios, and strategies for precise feeding amounts from feeding equipment in feeding scenarios.

[0072] In one possible implementation, the N scene control adjustment spaces are repeatedly optimized using a scene control coordination degree evaluation model to generate N scene control optimization strategies. Step S500 further includes step S510, which involves analyzing the scene control coordination degree of the nth scene control adjustment space based on the nth operation scene feature data and the scene control coordination degree evaluation model to obtain a scene control coordination degree evaluation set. Specifically, a neural network-based scene control coordination degree evaluation model is used to analyze the nth scene control adjustment space. The input data of this model includes the nth operation scene feature data (such as environmental data, animal status data, etc.) and control adjustment schemes (combinations of equipment operating parameters). For example, in the feed feeding equipment layer of a ranch, the operation scene feature data includes the hunger level of cattle and sheep, weather conditions, etc., and the control adjustment scheme is a combination of parameters such as the feeding frequency and feeding amount of the feeding equipment. The output of the model is the scene control coordination degree, which reflects the degree of matching between the control adjustment scheme and the operation scene, that is, whether the equipment operation mode can well adapt to the current scene requirements. By evaluating each control and adjustment scheme in the nth scenario control and adjustment space, a scenario control coordination evaluation set is obtained. This is a set of data containing different control and adjustment schemes and their corresponding coordination degrees.

[0073] Step S520: Based on the scene control coordination degree evaluation set, the nth scene control adjustment space is optimized and screened according to a predetermined scene control coordination degree to establish an initial optimization group for the nth scene control. Specifically, the scene control coordination degree evaluation set is screened according to a pre-set predetermined scene control coordination degree. The predetermined scene control coordination degree is a threshold value used to ensure that only solutions with sufficiently high coordination degrees are retained. For example, if the predetermined scene control coordination degree is set to 0.8, only control adjustment solutions with a coordination degree greater than or equal to 0.8 will be selected from the scene control coordination degree evaluation set of the milking equipment layer to establish the initial optimization group for the nth scene control. The solutions in this initial optimization group are all candidate solutions that perform well in terms of coordination degree.

[0074] Step S530: Based on the scene control coordination degree evaluation model and the predetermined scene control coordination degree, the initial optimization group of the nth scene control is subjected to breeding optimization to obtain the first breeding optimization group of scene control. Specifically, using breeding optimization algorithms such as genetic algorithms, combined with the scene control coordination degree evaluation model and the predetermined scene control coordination degree, the initial optimization group of the nth scene control is subjected to breeding optimization. In this process, the schemes in the initial optimization group will undergo genetic operations such as selection, crossover, and mutation to generate new schemes. For example, in the cattle shed cleaning equipment layer of a ranch, the selection operation retains the cleaning schemes with high coordination degree, the crossover operation combines the advantages of different schemes, and the mutation operation introduces new parameter changes, thereby forming the first breeding optimization group of scene control. These new schemes will be evaluated again by the scene control coordination degree evaluation model to ensure that their coordination degree still meets the predetermined requirements.

[0075] Step S540: Based on the scene control coordination degree evaluation model and the predetermined scene control coordination degree, continue to perform breeding optimization on the first scene control breeding optimization group to obtain the Qth scene control breeding optimization group, where Q is a positive integer greater than 1. Specifically, continue to use the breeding optimization algorithm and the scene control coordination degree evaluation model to perform multiple rounds of breeding optimization on the first scene control breeding optimization group to obtain the Qth scene control breeding optimization group (Q is a positive integer greater than 1). Each round of breeding optimization further optimizes the coordination degree of the schemes while maintaining a stable number of schemes. For example, in the automatic drinking water equipment layer of a pasture, after multiple rounds of breeding optimization, schemes with lower coordination degree are continuously eliminated, while schemes with higher coordination degree are retained and optimized, ultimately obtaining the Qth scene control breeding optimization group, in which the schemes perform better in adapting to the needs of the operational scene.

[0076] Step S550: Based on the initial optimization group of the nth scene control, the first reproductive optimization group of the scene control, ..., the Qth reproductive optimization group of the scene control, energy consumption minimization optimization is performed to generate the nth scene control optimization strategy. Specifically, after completing multiple rounds of reproductive optimization, all schemes from the initial optimization group of the nth scene control, the first reproductive optimization group of the scene control, up to the Qth reproductive optimization group of the scene control are integrated. Then, these schemes are further screened according to the goal of minimizing energy consumption. For example, in the lighting equipment layer of the ranch, the equipment energy consumption under each scheme is calculated, the scheme with the minimum energy consumption is found, and finally the nth scene control optimization strategy is generated. This optimization strategy is the control scheme that minimizes equipment energy consumption while satisfying scene control coordination.

[0077] This approach uses a scenario control coordination evaluation model to accurately assess and select control and adjustment schemes, ensuring that the final generated scenario control optimization strategy can well adapt to the needs of the operational scenario. This avoids problems caused by the incoordination between equipment operation and scenario requirements. The final energy consumption minimization optimization step can provide the ranch with the lowest energy consumption control strategy that meets the scenario control coordination requirements, which helps the ranch achieve its energy conservation and emission reduction goals and improve the ranch's economic benefits and environmental sustainability.

[0078] In one possible implementation, the initial optimization group for the nth scene control is subjected to reproductive optimization based on the scene control coordination degree evaluation model and the predetermined scene control coordination degree to obtain a first reproductive optimization group for scene control. Step S530 further includes step S531, allocating reproductive capacity to the initial optimization group for the nth scene control according to reproductive capacity constraints to obtain reproductive capacity allocation results. Specifically, reproductive capacity constraints refer to limiting the scale of reproductive operations to ensure the efficiency and feasibility of the optimization process. The system allocates reproductive capacity to the initial optimization group for the nth scene control according to reproductive capacity constraints. The reproductive capacity allocation is based on the magnitude of the scene control coordination degree, and the specific steps are as follows: For each scheme in the initial optimization group for the nth scene control, the weight is calculated according to its scene control coordination degree. A total reproductive capacity limit is set, and the allocatable reproductive capacity is calculated according to the weight of each scheme.

[0079] Step S532: Based on the breeding capacity allocation result, the nth scene control initial optimization group is bred according to the nth control trigger domain to obtain the first scene control breeding group. Specifically, based on the breeding capacity allocation result, the nth scene control initial optimization group is bred using the nth control trigger domain. The breeding operation includes genetic algorithm operations such as selection, crossover, and mutation. The specific steps are as follows: According to the breeding capacity allocation result, a scheme is selected from the nth scene control initial optimization group for breeding. The number of times each scheme is selected is equal to its allocated breeding quantity. The selected schemes are paired up and crossover is performed to generate new offspring schemes. The crossover operation can use single-point crossover or multi-point crossover methods. The offspring schemes are mutated to increase the diversity of the schemes. The mutation operation can use methods such as randomly changing certain control parameters.

[0080] Step S533: Perform multi-dimensional risk joint optimization on the first breeding population under scenario control based on multi-dimensional risk constraints to obtain the first breeding optimization population under scenario control. Specifically, perform multi-dimensional risk joint optimization on the schemes in the first breeding population under scenario control, and use multi-dimensional risk constraints (scenario control mismatch risk constraint, induced livestock risk constraint, and equipment failure risk constraint) to conduct risk assessment on each scheme through a risk assessment model. The specific steps are as follows: For each offspring scheme, predict its scenario control mismatch risk coefficient, induced livestock risk coefficient, and equipment failure risk coefficient. Based on the multi-dimensional risk constraints, select the schemes that satisfy all risk constraints to obtain the first breeding optimization population under scenario control.

[0081] Step S534: Based on the scene control coordination degree evaluation model, optimize the scene control coordination degree of the first breeding optimization group according to the predetermined scene control coordination degree to generate the first breeding optimization group of scene control. Specifically, using the scene control coordination degree evaluation model, evaluate the scene control coordination degree of the schemes in the first breeding optimization group of scene control. According to the predetermined scene control coordination degree, select schemes with a coordination degree greater than or equal to the predetermined value to generate the first breeding optimization group of scene control.

[0082] This implementation method reasonably limits the breeding operation through breeding capacity constraints, avoiding the waste of computational resources and excessively lengthy optimization processes caused by excessively large breeding scales. This makes the optimization process more efficient, generating high-quality offspring solutions within limited resources and time. Introducing a multi-dimensional risk joint optimization step during the breeding process ensures that the newly generated solutions meet constraints regarding risks such as scenario control mismatch, induced livestock and equipment failures. This makes the final scenario control optimization strategy safer and more reliable, reducing potential risks in practical applications and ensuring the stable operation of ranch equipment and the safety of livestock. Through scenario control coordination optimization, highly adaptable solutions are further selected. These solutions better match the actual operational needs of the ranch, improving equipment operating efficiency and production benefits.

[0083] Step S600: Conflict optimization is performed based on the N scene control optimization strategies to obtain a global control optimization strategy, and the N ranch equipment layers are centrally controlled based on the global control optimization strategy.

[0084] Specifically, the global control optimization strategy refers to the optimal strategy applicable to the centralized control of the entire ranch equipment, obtained by comprehensively considering the control optimization strategies of each scenario and resolving the conflicts between them, which can achieve efficient and stable operation of the ranch equipment.

[0085] Using a conflict graph model from graph theory, conflicts between control optimization strategies in different scenarios are represented as edges in the graph, with nodes representing each optimization strategy. Then, a coloring algorithm is applied to color the conflict graph, identifying non-conflicting strategy combinations to generate a global control optimization strategy. Finally, using control technologies such as a distributed control system (DCS) or a programmable logic controller (PLC), control commands are sent to the devices at N ranch equipment layers according to the global control optimization strategy, achieving centralized control.

[0086] For example, these N scenario control optimization strategies can be represented by a conflict graph. For instance, there may be edges between the feeding strategy and the cattle shed environment control strategy due to conflicts in equipment power load. A coloring algorithm is used to determine the global control optimization strategy without conflicts, that is, each scenario device operates with optimized parameters at the appropriate time. Finally, the DCS system sends control signals to each device according to this global strategy to achieve centralized control of the equipment.

[0087] This application embodiment identifies work scenarios using a multi-source sensing dataset, obtaining N work scenario feature data. Based on these feature data, the ranch equipment set is adaptively classified to form N corresponding ranch equipment layers. Based on the work scenario feature data, the current control strategy of each equipment layer is risk-predicted, generating N risk prediction sequences. These sequences are used to adjust the control strategy for risk suppression, constructing N scenario control adjustment spaces. The adjustment spaces are then repeatedly optimized using a scenario control coordination degree evaluation model to generate N scenario control optimization strategies. These optimization strategies are then conflict-optimized and integrated to generate a global control optimization strategy. Based on this, centralized control is implemented for all ranch equipment layers. These technical means solve the technical problem of insufficient control flexibility and effectiveness in existing centralized control of ranch equipment, achieving the technical effect of improving control flexibility and effectiveness.

[0088] In the above text, refer to Figure 1 A scenario-adaptive centralized control method for ranch equipment according to an embodiment of the present invention is described in detail. Next, reference will be made to... Figure 2 A scenario-adaptive centralized control system for ranch equipment is described according to an embodiment of the present invention.

[0089] An adaptive ranch equipment centralized control system according to an embodiment of the present invention addresses the technical problem of insufficient control flexibility and effectiveness in existing centralized ranch equipment control systems, thereby improving the technical effect of control flexibility and effectiveness. The adaptive ranch equipment centralized control system includes: a work scene identification module 10, an equipment adaptive classification module 20, a control strategy risk prediction module 30, a risk suppression and adjustment module 40, an optimization module 50, and a conflict optimization module 60.

[0090] The system comprises the following modules: a task scene identification module 10, which identifies task scenes based on a multi-source perception dataset of the target ranch and obtains N task scene feature data, where N is a positive integer greater than 1; an equipment adaptive classification module 20, which adaptively classifies the ranch equipment set of the target ranch based on the N task scene feature data and obtains N ranch equipment layers; a control strategy risk prediction module 30, which predicts the risks of N current control strategies for the N ranch equipment layers based on the N task scene feature data and obtains N risk prediction sequences; a risk suppression and adjustment module 40, which performs risk suppression and adjustment on the N current control strategies based on the N risk prediction sequences and generates N scenario control adjustment spaces; an optimization module 50, which performs multiple breeding optimizations on the N scenario control adjustment spaces based on a scenario control coordination degree evaluation model and generates N scenario control optimization strategies; and a conflict optimization module 60, which performs conflict optimization based on the N scenario control optimization strategies to obtain a global control optimization strategy and performs centralized control on the N ranch equipment layers based on the global control optimization strategy.

[0091] The specific configuration of the control strategy risk prediction module 30 will be described in detail below. As mentioned above, based on the N operational scenario feature data, risk prediction is performed on the N current control strategies of the N ranch equipment layers to obtain N risk prediction sequences. The control strategy risk prediction module 30 may further include: a data extraction unit used to extract the nth operational scenario feature data and the nth current control strategy corresponding to the nth ranch equipment layer based on the N operational scenario feature data and the N current control strategies, where n is a positive integer, 1≤n≤N; and a scenario control mismatch risk prediction unit used to predict the risk of the nth current control strategy based on the nth operational scenario feature data. The strategy performs scenario control mismatch risk prediction to obtain the nth scenario control mismatch risk coefficient; the livestock induction risk prediction unit is used to perform livestock induction risk prediction on the nth current control strategy based on the nth operation scenario feature data to obtain the nth livestock induction risk coefficient; the equipment failure risk prediction unit is used to perform equipment failure risk prediction on the nth current control strategy based on the nth operation scenario feature data to obtain the nth equipment failure risk coefficient, and combine the nth scenario control mismatch risk coefficient and the nth livestock induction risk coefficient to generate the nth risk prediction sequence.

[0092] Specifically, based on the feature data of the nth operation scenario, the nth current control strategy is used to predict the scenario control mismatch risk, and the nth scenario control mismatch risk coefficient is obtained. The scenario control mismatch risk prediction unit may further include: a scenario control mismatch risk record retrieval subunit for retrieving scenario control mismatch risk records of the target ranch to obtain a control strategy sample set and a scenario control mismatch risk sample set; a clustering subunit for clustering the scenario control mismatch risk sample set according to the control strategy sample set to obtain each scenario control mismatch risk group corresponding to each control strategy sample; and a confidence fusion subunit for performing confidence fusion on each scenario control mismatch risk group. The system combines the control strategy sample set and the mismatch risk sample confidence set to obtain a risk prediction learner supervised training subunit. The risk prediction learner is trained in a supervised manner based on the control strategy sample set and the mismatch risk sample confidence set, and a mismatch risk prediction loss coefficient is obtained after each predetermined number of training iterations. The scenario control mismatch risk prediction model acquisition subunit is used to obtain a scenario control mismatch risk prediction model if the mismatch risk prediction loss coefficient is less than the mismatch risk prediction loss threshold. The scenario control mismatch risk coefficient acquisition subunit is used to input the nth operation scenario feature data and the nth current control strategy into the scenario control mismatch risk prediction model to obtain the nth scenario control mismatch risk coefficient.

[0093] The specific configuration of the risk suppression and adjustment module 40 will be described in detail below. As mentioned above, risk suppression and adjustment are performed on the N current control strategies based on the N risk prediction sequences to generate N scenario control adjustment spaces. The risk suppression and adjustment module 40 may further include: a judgment unit for judging whether the nth risk prediction sequence satisfies multivariate risk constraints, which include scenario control mismatch risk constraints, livestock induction risk constraints, and equipment failure risk constraints; a control scheme retrieval unit for retrieving control schemes for the nth pasture equipment layer based on the nth operation scenario feature data if the nth risk prediction sequence does not satisfy the multivariate risk constraints, and establishing an nth control scheme retrieval set; a multidimensional control trigger feature parsing unit for performing multidimensional control trigger feature parsing on the nth pasture equipment layer based on the nth control scheme retrieval set to obtain the nth control trigger domain; a multidimensional adjustment unit for performing multidimensional adjustment on the nth current control strategy based on the nth control trigger domain to obtain the nth control adjustment group; and a multidimensional risk joint optimization unit for performing multidimensional risk joint optimization on the nth control adjustment group based on the multivariate risk constraints to generate the nth scenario control adjustment space.

[0094] Specifically, the control and regulation group n is subjected to multi-dimensional risk joint optimization based on the multi-dimensional risk constraints to generate a control and regulation space for the nth scenario. The multi-dimensional risk joint optimization unit may further include: a control and regulation first scheme extraction subunit for extracting a control and regulation first scheme based on the control and regulation group n; a multi-dimensional risk prediction subunit for performing multi-dimensional risk prediction on the control and regulation first scheme based on the feature data of the nth operation scenario to obtain a first scheme risk prediction sequence; a judgment subunit for judging whether the first scheme risk prediction sequence satisfies the multi-dimensional risk constraints; and an nth scenario control and regulation space generation subunit for adding the control and regulation first scheme to the nth scenario control and regulation space if the first scheme risk prediction sequence satisfies the multi-dimensional risk constraints.

[0095] The specific configuration of the optimization module 50 will be described in detail below. As mentioned above, based on the scene control coordination degree evaluation model, the N scene control adjustment spaces are subjected to multiple breeding optimizations to generate N scene control optimization strategies. The optimization module 50 may further include: a scene control coordination degree analysis unit used to analyze the scene control coordination degree of the nth scene control adjustment space based on the nth job scene feature data and the scene control coordination degree evaluation model to obtain a scene control coordination degree evaluation set; an optimization screening unit used to perform optimization screening on the nth scene control adjustment space based on the scene control coordination degree evaluation set and a predetermined scene control coordination degree to establish an initial optimization group for the nth scene control; and a breeding optimization unit used to... The initial optimization group for the nth scene control is subjected to reproductive optimization based on the scene control coordination degree evaluation model and the predetermined scene control coordination degree to obtain the first reproductive optimization group for scene control. The reproductive optimization iteration unit is used to continue to perform reproductive optimization on the first reproductive optimization group for scene control based on the scene control coordination degree evaluation model and the predetermined scene control coordination degree to obtain the Qth reproductive optimization group for scene control, where Q is a positive integer greater than 1. The energy consumption minimization optimization unit is used to perform energy consumption minimization optimization based on the initial optimization group for the nth scene control, the first reproductive optimization group for scene control, ..., the Qth reproductive optimization group for scene control to generate the optimization strategy for the nth scene control.

[0096] Specifically, the initial optimization group of the nth scene control is subjected to reproductive optimization based on the scene control coordination degree evaluation model and the predetermined scene control coordination degree to obtain a first reproductive optimization group of scene control. The reproductive optimization unit may further include: a reproductive capacity allocation subunit for allocating reproductive capacity to the initial optimization group of the nth scene control according to reproductive capacity constraints to obtain a reproductive capacity allocation result; a reproductive subunit for reproductively optimizing the initial optimization group of the nth scene control according to the nth control trigger domain based on the reproductive capacity allocation result to obtain a first reproductive group of scene control; a multi-dimensional risk joint optimization subunit for performing multi-dimensional risk joint optimization on the first reproductive group of scene control according to multi-dimensional risk constraints to obtain a first reproductive optimization group of scene control; and a scene control coordination degree optimization subunit for performing scene control coordination optimization on the first reproductive optimization group of scene control according to the predetermined scene control coordination degree based on the scene control coordination degree evaluation model to generate the first reproductive optimization group of scene control.

[0097] The specific configuration of the operation scene recognition module 10 will be described in detail below. As mentioned above, the operation scene recognition module 10 can further include: a multi-source monitoring unit for performing multi-source monitoring on the target pasture to obtain a pasture monitoring dataset; and a data cleaning unit for cleaning the pasture monitoring dataset to generate the multi-source sensing dataset.

[0098] The specific configuration of the device adaptive classification module 20 will be described in detail below. As mentioned above, to obtain N ranch equipment layers, the device adaptive classification module 20 may further include: a real-time control parameter acquisition unit for acquiring real-time control parameters of each device in the ranch equipment set to obtain a device control dataset; and a clustering unit for clustering the device control dataset according to the N ranch equipment layers to generate the N current control strategies.

[0099] The scene-adaptive centralized control system for ranch equipment provided in this embodiment of the invention can execute the scene-adaptive centralized control method for ranch equipment provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0100] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0101] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A scenario-adaptive centralized control method for ranch equipment, characterized in that, The method includes: Based on the multi-source perception dataset of the target ranch, the operation scene is identified to obtain N operation scene feature data, where N is a positive integer greater than 1; Based on the N operational scenario feature data, the ranch equipment set of the target ranch is adaptively classified to obtain N ranch equipment layers; Based on the characteristic data of the N operational scenarios, risk prediction is performed on the N current control strategies of the N ranch equipment layers to obtain N risk prediction sequences; Based on the N risk prediction sequences, risk suppression and adjustment are performed on the N current control strategies to generate N scenario control adjustment spaces; Based on the scene control coordination degree evaluation model, the N scene control adjustment spaces are subjected to multiple breeding optimizations to generate N scene control optimization strategies. Conflict optimization is performed based on the N scenario control optimization strategies to obtain a global control optimization strategy, and the N ranch equipment layers are centrally controlled based on the global control optimization strategy.

2. The scene-adaptive centralized control method for ranch equipment as described in claim 1, characterized in that, Based on the characteristic data of the N operational scenarios, risk prediction is performed on the N current control strategies of the N ranch equipment layers to obtain N risk prediction sequences, including: Based on the N operational scenario feature data and the N current control strategies, extract the nth operational scenario feature data and the nth current control strategy corresponding to the nth pasture equipment layer, where n is a positive integer and 1≤n≤N; Based on the feature data of the nth operation scenario, the scenario control mismatch risk is predicted for the nth current control strategy to obtain the scenario control mismatch risk coefficient. Based on the characteristic data of the nth operation scenario, the risk of livestock induced by the nth current control strategy is predicted to obtain the risk coefficient of livestock induced by the nth operation scenario. Based on the feature data of the nth operation scenario, the equipment failure risk is predicted for the nth current control strategy to obtain the nth equipment failure risk coefficient. Combined with the control mismatch risk coefficient of the nth scenario and the induced livestock risk coefficient, the nth risk prediction sequence is generated.

3. The scene-adaptive centralized control method for ranch equipment as described in claim 2, characterized in that, Based on the characteristic data of the nth operation scenario, the scenario control mismatch risk is predicted for the nth current control strategy to obtain the scenario control mismatch risk coefficient, including: The target ranch is subjected to a scene control mismatch risk record retrieval to obtain a control strategy sample set and a scene control mismatch risk sample set. Cluster the scenario control mismatch risk sample set based on the control strategy sample set to obtain the scenario control mismatch risk group corresponding to each control strategy sample; Confidence fusion is performed on the control mismatch risk groups of each scenario to obtain a confidence set of mismatch risk samples; The risk prediction learner is trained under supervision based on the control strategy sample set and the mismatch risk sample confidence set. After each predetermined number of training iterations, the mismatch risk prediction loss coefficient is obtained. If the mismatch risk prediction loss coefficient is less than the mismatch risk prediction loss threshold, the scenario control mismatch risk prediction model is obtained. The nth operation scenario feature data and the nth current control strategy are input into the scenario control mismatch risk prediction model to obtain the nth scenario control mismatch risk coefficient.

4. The scene-adaptive centralized control method for ranch equipment as described in claim 1, characterized in that, Based on the N risk prediction sequences, risk suppression adjustments are made to the N current control strategies to generate N scenario control adjustment spaces, including: Determine whether the nth risk prediction sequence satisfies the multivariate risk constraints, which include scenario control mismatch risk constraints, livestock induced risk constraints, and equipment failure risk constraints. If the nth risk prediction sequence does not meet the multivariate risk constraints, control scheme retrieval is performed on the nth pasture equipment layer based on the nth operation scenario feature data to establish the nth control scheme retrieval set. Based on the nth control scheme retrieval set, multi-dimensional control trigger feature analysis is performed on the nth pasture equipment layer to obtain the nth control trigger domain; Based on the nth control trigger domain, the nth current control strategy is adjusted in multiple dimensions to obtain the nth control adjustment group; Based on the aforementioned multivariate risk constraints, a multidimensional risk joint optimization is performed on the nth control and regulation group to generate the nth scenario control and regulation space.

5. The scene-adaptive centralized control method for ranch equipment as described in claim 4, characterized in that, Based on the aforementioned multivariate risk constraints, a multidimensional risk joint optimization is performed on the nth group of control and regulation to generate the nth scenario control and regulation space, including: Based on the nth control and regulation group, extract the first control and regulation scheme; Based on the feature data of the nth operation scenario, multidimensional risk prediction is performed on the first control and adjustment scheme to obtain the risk prediction sequence of the first scheme; Determine whether the risk prediction sequence of the first scheme satisfies the multivariate risk constraints; If the risk prediction sequence of the first scheme satisfies the multivariate risk constraints, the first control and regulation scheme is added to the control and regulation space of the nth scenario.

6. The scene-adaptive centralized control method for ranch equipment as described in claim 1, characterized in that, Based on the scene control coordination evaluation model, the N scene control adjustment spaces are subjected to multiple iterations of optimization to generate N scene control optimization strategies, including: Based on the feature data of the nth operation scenario, and according to the scenario control coordination degree evaluation model, the scenario control adjustment space of the nth scenario is analyzed to obtain the scenario control coordination degree evaluation set. Based on the scene control coordination degree evaluation set, the nth scene control adjustment space is optimized and screened according to the predetermined scene control coordination degree to establish the initial optimization group of the nth scene control. Based on the scene control coordination degree evaluation model and the predetermined scene control coordination degree, the initial optimization group of the nth scene control is subjected to breeding optimization to obtain the first breeding optimization group of scene control. Based on the scene control coordination degree evaluation model and the predetermined scene control coordination degree, the first scene control breeding optimization group is further bred to obtain the Qth scene control breeding optimization group, where Q is a positive integer greater than 1; Based on the initial optimization group of the nth scenario control, the first reproductive optimization group of the scenario control, ... the Qth reproductive optimization group of the scenario control, energy consumption minimization optimization is performed to generate the nth scenario control optimization strategy.

7. The scene-adaptive centralized control method for ranch equipment as described in claim 6, characterized in that, Based on the scene control coordination degree evaluation model and the predetermined scene control coordination degree, the initial optimization group of the nth scene control is subjected to breeding optimization to obtain the first breeding optimization group of scene control, including: Based on the reproductive capacity constraint, the initial optimization group of the nth scenario is controlled to allocate the reproductive capacity, and the reproductive capacity allocation result is obtained. Based on the breeding capacity allocation result, the nth scene control initial optimization group is bred according to the nth control trigger domain to obtain the scene control first breeding group; Based on multiple risk constraints, the first breeding group of the scenario control is subjected to multi-dimensional risk joint optimization to obtain the first breeding optimization group of the scenario control. Based on the scene control coordination degree evaluation model, the scene control coordination degree of the first breeding optimization group is optimized according to the predetermined scene control coordination degree to generate the first breeding optimization group of scene control.

8. The scene-adaptive centralized control method for ranch equipment as described in claim 1, characterized in that, Operational scene identification is performed based on the multi-source sensing dataset of the target pasture, including: Multi-source monitoring was performed on the target pasture to obtain a pasture monitoring dataset; The ranch monitoring dataset is cleaned to generate the multi-source sensing dataset.

9. The scene-adaptive centralized control method for ranch equipment as described in claim 1, characterized in that, Obtain N ranch equipment layers, including: Collect real-time control parameters for each device within the ranch equipment set to obtain an equipment control dataset; The equipment control dataset is clustered based on the N ranch equipment layers to generate the N current control strategies.

10. A scene-adaptive centralized control system for ranch equipment, characterized in that, The system is used to implement the scene-adaptive centralized control method for ranch equipment as described in any one of claims 1-9, the system comprising: The operation scene recognition module is used to identify operation scenes based on the multi-source perception dataset of the target ranch and obtain N operation scene feature data, where N is a positive integer greater than 1. The equipment adaptive classification module is used to adaptively classify the set of ranch equipment in the target ranch based on the N operational scenario feature data to obtain N ranch equipment layers; The control strategy risk prediction module is used to predict the risk of N current control strategies of the N ranch equipment layers based on the N operational scenario feature data, and obtain N risk prediction sequences. The risk suppression and adjustment module is used to perform risk suppression and adjustment on the N current control strategies based on the N risk prediction sequences, and generate N scenario control adjustment spaces. The optimization module is used to perform multiple breeding optimizations on the N scene control adjustment spaces according to the scene control coordination degree evaluation model to generate N scene control optimization strategies. The conflict optimization module is used to perform conflict optimization based on the N scenario control optimization strategies, obtain a global control optimization strategy, and perform centralized control on the N ranch equipment layers based on the global control optimization strategy.

Images