Breeding environment management system and method applied to smart dairy farm

Abstract

The present application relates to the field of livestock breeding technology, especially to a breeding environment management and control system and method applied to a smart dairy farm, specifically comprising a breeding environment management and control center, an environment sensing data module, a model and influence management module, a scene recognition module, a scene matching and regulation module, a dynamic emergency management module and a visual board module; through standardized collection and processing of multi-source heterogeneous data, the data foundation of intelligent management and control is consolidated, a differentiated environment-performance regression correlation model is constructed for different physiological stages of dairy cows, accurate quantitative prediction of dairy cow performance and effective identification of environmental risk factors are realized, through multi-feature fusion scene recognition, the breeding scene is accurately determined, and personalized environment regulation strategies oriented to dairy cow performance recovery are matched, a dynamic closed-loop management and control mechanism and standardized equipment fault replacement screening rules are established, the gap in management and control capacity is quickly filled, so as to realize the management effect of dairy cow environment regulation.

Description

Technical Field

[0001] This invention relates to the field of livestock farming technology, and in particular to a farming environment management system and method for use in smart dairy farms. Background Technology

[0002] With the development of large-scale and intensive animal husbandry, the dairy farming industry is gradually transforming from traditional manual experience-based management to intelligent and data-driven management. The growth, milk production, and other production performance of dairy cows are highly correlated with the barn environment. Environmental parameters such as temperature and humidity, ammonia / carbon dioxide concentration, dust content, wind speed, and light in the barn directly affect the feed intake, rumination efficiency, physical health, and milk production performance of dairy cows. Therefore, precise control of the barn environment is the core link to ensure the production performance of dairy cows and improve farming efficiency.

[0003] To improve the cattle shed environment, environmental sensors for temperature, humidity, and ammonia, as well as environmental control equipment such as fans, wet curtains, and sprinklers, have been gradually introduced to achieve real-time collection of some environmental parameters and semi-automatic control of the equipment. Some farms have also collected basic physiological data of dairy cows by using devices such as electronic ear tags and smart collars, attempting to combine environmental data with dairy cow physiological data for breeding management. This has alleviated to some extent the problems of low efficiency and strong subjectivity of traditional manual inspection and manual control.

[0004] However, in actual large-scale farming applications, there are still many shortcomings in the technology for controlling the environment of cattle sheds and managing the performance of dairy cows, making it difficult to meet the needs of refined and intelligent farming. The specific manifestations are as follows: Data collection and processing lack systematicity: multi-source heterogeneous data is not collected in an integrated manner, which cannot support the correlation analysis between the environment and dairy cow performance. Moreover, the management and control strategies lack precision and adaptability. Differentiated control schemes are not designed for different physiological stages of dairy cows. The control is only aimed at achieving environmental parameters, which is out of touch with dairy cow performance. Furthermore, there is no ability to identify complex scenarios, and the strategies are not targeted enough. 2. Imperfect control mechanism and low level of visualization: There is no dynamic closed-loop feedback optimization mechanism, and various control data are displayed in a scattered manner. There is no integrated visualization system, resulting in low management efficiency and delayed decision-making.

[0005] In summary, this paper proposes an intelligent management method for the environment of cattle sheds and the performance of dairy cows, addressing the aforementioned technical deficiencies. This method aims to improve the accuracy and intelligence of cattle shed environment management, achieve coordinated management of environmental control and dairy cow production performance, and ultimately improve farming efficiency. Summary of the Invention

[0006] The purpose of this invention is to provide a breeding environment management system and method for smart dairy farms, in order to solve the aforementioned technical defects.

[0007] The objective of this invention can be achieved through the following technical solution: a breeding environment management and control system applied to smart dairy farms, including a breeding environment management and control center, an environmental perception data module, a model and impact management module, a scene recognition module, a scene matching and control module, a dynamic emergency management module, and a visual dashboard module; The environmental perception data module is used to collect environmental parameters, regional image data, physiological parameters of dairy cows, and equipment status parameters of cattle sheds, construct multi-source heterogeneous data, preprocess the multi-source heterogeneous data to form standardized multi-dimensional raw data, and send it to the breeding environment management and control center for storage. The Model and Impact Management module is used to construct environmental-performance sample datasets for each physiological stage, and to construct a regression correlation model of environmental-cow performance for multiple physiological stages. It also judges the quantitative prediction results of output environmental parameters and cow performance, and outputs performance impact risk factors. The scene recognition module is used to extract features and perform scene discrimination analysis on standardized multidimensional raw data. It compares the obtained current environment state vector one by one, and if triggered, it outputs control instructions. The scene matching and control module is used to respond to control commands and, in conjunction with sliding window calculation, generate the fusion state vector Vt at the current moment. Based on the analysis of the fusion state vector Vt and combined with the preset scene-control mapping rule library, it outputs the environmental control strategy and the expected performance recovery target, and generates executable control commands. After the control commands corresponding to the environmental control strategy are issued to the environmental control equipment, the dynamic emergency management module activates the real-time data acquisition and feedback mechanism.

[0008] Preferably, the standardized multidimensional raw data analysis process is as follows: Collect environmental parameters, regional image data, dairy cow physiological parameters, and equipment status parameters of cattle sheds to construct multi-source heterogeneous data; The collected multi-source heterogeneous data is preprocessed to form standardized multidimensional raw data.

[0009] Preferably, the analysis process for the performance impact risk factors is as follows: Standardized multidimensional raw data were stratified and filtered according to the physiological stage labels of dairy cows to construct environmental-performance sample datasets for each physiological stage. Based on the environment-performance sample datasets of each physiological stage, a regression correlation model of environment-dairy cow performance in multiple physiological stages is constructed. Real-time environmental parameters are input into the environment-dairy cow performance regression correlation model corresponding to the physiological stage, and the output is the quantitative prediction result of environmental parameters-dairy cow performance, that is, the predicted value of the overall growth and milk production performance of dairy cows in the current environment; at the same time, combined with the pre-set farm dairy cow breed benchmark performance value and high-quality production performance range, the critical warning value of environmental parameters is calibrated - when a change in a certain environmental parameter causes the performance prediction value to be lower than a% of the benchmark value (a>0), or to exceed the high-quality performance range, the environmental parameter is marked as a performance impact risk factor.

[0010] Preferably, the feature extraction and scene discrimination analysis process is as follows: Based on standardized multidimensional raw data, environmental features, cattle status features, equipment operating condition features, and performance correlation features are extracted. Based on the physiological stage labels of the dairy cows in the barn, the basic comfort threshold vector corresponding to the physiological stage labels is retrieved. At the same time, the critical warning values ​​of environmental parameters calibrated by the environment-dairy cow performance correlation model are retrieved. These are fused to construct a performance-oriented current environmental state vector. The current environmental state vector is then compared one by one with the basic comfort threshold vector. If any environmental feature (such as dry-bulb temperature) in the current environmental state vector exceeds the threshold in the corresponding basic comfort threshold vector, or if any performance-related feature triggers a risk warning, then a control command will be triggered.

[0011] Preferably, the analysis process of the fused state vector Vt is as follows: S1: When generating control commands, based on environmental characteristics, cattle status characteristics, and equipment operating condition characteristics, and combined with sliding window calculations, the current fusion state vector Vt = [Et, Ct, Mt, Pt] is generated, where t is greater than zero and t represents the sampling time; Where Et is the environmental feature vector, Ct is the cattle state feature vector, Mt is the equipment operating condition feature vector, and Pt is the performance correlation feature vector; S2: Preprocess the constructed fusion state vector Vt and extract temporal extended features to form an extended feature vector KVt.

[0012] Preferred options also include: S3: Retrieve the pre-built scene template library and obtain the scene template Sk for each scene in the scene template library; S4: Calculate the matching degree between the current extended feature vector KVt and each scene template Sk one by one; The similarity of each feature is weighted and fused according to the feature weight vector Wk to obtain the comprehensive scene confidence Zxk; S5: Filter out all scene templates with a comprehensive scene confidence score Zxk≥θk as candidate matching scenes; If there is only one candidate scenario, it is directly determined as the current single scenario label; if there are multiple candidate scenarios, arbitration logic is executed. Priority takes precedence: the scene with the highest priority in PyK is selected as the main scene; Confidence-weighted fusion: If multiple high-priority scenarios coexist, a composite scenario label is generated; S6: Based on the preset scene-control mapping rule library, match the environmental control strategy and expected performance recovery target corresponding to the scene label, and generate executable control instructions.

[0013] Preferably, the analysis process of the real-time data acquisition and feedback mechanism is as follows: The environmental perception data module continuously collects dynamic data on the characteristics of the cowshed environment, the state of the dairy cows, and the operating conditions of the equipment. At the same time, it collects real-time data related to the performance of the dairy cows, generates a real-time feature vector ZVt after execution, and compares the difference with the fused state vector Vt before the instruction is executed to obtain the feature change difference ΔV. Set a threshold range [ΔVmin, ΔVmax] for judging the effect of regulation: If ΔV falls into the threshold range and the change value of the performance-related feature reaches the expected performance recovery target, the regulation is judged to be effective; if ΔV < ΔVmin, or the change value of the performance-related feature does not reach the expected performance recovery target, the regulation is judged to be insufficient; if ΔV > ΔVmax, the regulation is judged to be excessive. If an abnormal operating condition is detected when the equipment executes instructions, the current control instructions will be immediately suspended, and equipment fault warning and replacement analysis will be triggered.

[0014] Preferably, the equipment fault early warning and replacement analysis process is as follows: Obtain the unique identifier and installation area of ​​the faulty equipment, retrieve the equipment feature database, obtain the core control capabilities of the faulty equipment, and at the same time, identify the control capability gap in the area after the faulty equipment fails. Match alternative devices according to the set filtering rules.

[0015] The beneficial effects of this invention are as follows: This invention strengthens the foundation of management data by collecting and standardizing multi-source data from all dimensions, thus solving the problems of fragmented and inconsistent data collection in traditional animal husbandry. At the same time, it constructs a hierarchical environment-performance sample dataset and a differentiated regression correlation model based on the growth and milk production characteristics of dairy cows at different physiological stages. This allows for accurate quantitative prediction of dairy cow performance under the current environment, rapid identification of environmental risk factors that lead to performance impairment, and early avoidance of the negative impact of environmental factors on dairy cow production performance.

[0016] This invention also accurately determines the current breeding scene in the cowshed through multi-feature fusion scene recognition and matching, and matches corresponding personalized control strategies. At the same time, it takes performance recovery as the control guide, so that the control commands are no longer limited to meeting environmental parameters, but focus more on ensuring and restoring the production performance of dairy cows. By collecting feedback data in real time, it compares and analyzes the feature changes before and after control, and judges the control effect by combining the physiological stage-specific threshold range. Differentiated strategies are implemented for three situations: effective control, insufficient control, and over-control. At the same time, standardized replacement screening rules are formulated for equipment failure, and equivalent replacement equipment is quickly matched and operating parameters are adjusted, effectively filling the gap in control capabilities and avoiding environmental loss of control due to equipment failure. This solves the problems of traditional control commands ending as soon as they are issued, no feedback optimization, and delayed response to equipment failure. Attached Figure Description

[0017] The invention will now be further described with reference to the accompanying drawings; Figure 1 This is a flowchart of the system of the present invention; Figure 2 This is a reference diagram for analyzing the control commands of this invention; Figure 3 This is a schematic diagram illustrating the method of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments; Example 1: Please refer to Figures 1 to 3As shown, this invention is a breeding environment control system applied to a smart dairy farm, including a breeding environment control center, an environmental perception data module, a model and impact management module, a scene recognition module, a scene matching and control module, a dynamic emergency management module, and a visual dashboard module. The environmental perception data module has a one-way communication connection with the breeding environment control center, the breeding environment control center has a two-way communication connection with the model and impact management module, the breeding environment control center has a one-way communication connection with the scene recognition module, the scene recognition module has a one-way communication connection with the scene matching and control module, the scene matching and control module has a one-way communication connection with both the dynamic emergency management module and the visual dashboard module, and the dynamic emergency management module has a one-way communication connection with the visual dashboard module. The environmental perception data module is used to collect environmental parameters, regional image data, physiological parameters of dairy cows, and equipment status parameters of cattle sheds, construct multi-source heterogeneous data, preprocess the multi-source heterogeneous data to form standardized multi-dimensional raw data, and send it to the breeding environment management and control center for storage. Among them, temperature and humidity sensors, ammonia sensors, carbon dioxide sensors, dust (PM2.5 / PM10) sensors, wind speed sensors and light intensity sensors are used to collect environmental parameters of the cattle shed in real time. High-definition network cameras deployed in the cattle sheds are used to collect images of the dairy cows' behavior, body condition, and feeding and drinking. The cow's physiological data, such as body temperature, activity level and rumination time, are collected in real time by electronic ear tags worn on the cow's ears or smart collars worn around its neck. The smart collars are equipped with temperature sensors, pedometers and accelerometers. Equipment status parameters: used to collect real-time operating status (such as start / stop status, current, voltage) and energy consumption data of environmental control equipment such as fans, wet curtains, sprinklers, roller shutters, and heaters; The collected multi-source heterogeneous data is preprocessed (such as data cleaning and noise reduction, data labeling, etc.) to form standardized multi-dimensional raw data, which is then sent to the aquaculture environment management and control center for storage. For example: data cleaning and noise reduction: removing glitch data caused by sensor malfunctions (such as instantaneous temperature jump of 20°C), blurry frames in video images, and invalid physiological data caused by collar detachment; Data labeling: The dataset is labeled. For example, video clips showing cows breathing with their mouths open and their breathing rate increasing are labeled as "heat stress state"; periods when milk production drops by more than 10% for three consecutive days are labeled as "impaired production performance"; and periods when the fans are running at full capacity but the temperature and humidity do not decrease are labeled as "insufficient equipment efficiency". These labels will serve as target variables for subsequent supervised learning. The Model and Impact Management module is used to construct environment-performance sample datasets for each physiological stage, and to build a multi-physiological-stage environment-cow performance regression correlation model. It then evaluates the quantitative prediction results of the output environment parameters and cow performance, and outputs performance impact risk factors, specifically including: Dairy cows are grouped and housed separately according to their physiological stage labels (such as peak lactation, calf stage, and growing stage). That is, dairy cows in peak lactation are grouped together and housed in one barn, dairy cows in the calf stage are grouped together and housed in one barn, etc. Standardized multidimensional raw data were stratified and filtered according to the physiological stage labels of dairy cows to construct environmental-performance sample datasets for each physiological stage, namely, environmental-performance sample datasets of cattle sheds for each physiological stage. Based on the existing Gradient Boosting Regression Tree (GBRT) combined with the Random Forest algorithm, a regression correlation model of environment-dairy cow performance at multiple physiological stages is constructed. Differentiated model training parameters are set for the overall growth and milk production characteristics of dairy cows at the same physiological stage in the same barn. Real-time environmental parameters are input into the environment-dairy cow performance regression correlation model corresponding to the physiological stage, and the output is the quantitative prediction result of environmental parameters-dairy cow performance, that is, the predicted value of the overall growth and milk production performance of dairy cows in the current environment; at the same time, combined with the pre-set farm dairy cow breed benchmark performance value and high-quality production performance range, the critical warning value of environmental parameters is calibrated - when a change in a certain environmental parameter causes the performance prediction value to be lower than a% of the benchmark value (a>0), or to exceed the high-quality performance range, the environmental parameter is marked as a performance impact risk factor.

[0020] Example 2: The scene recognition module is used to extract features and perform scene discrimination analysis on standardized multidimensional raw data. It compares the obtained current environment state vector one by one. If triggered, it outputs control instructions, specifically including: Based on standardized multidimensional raw data, environmental features, cattle status features, equipment operating condition features, and performance correlation features are extracted. Environmental feature extraction: including dry bulb temperature, relative humidity, ammonia concentration, etc.; Cattle condition characteristics extraction: rumination time, activity level (steps), percentage of resting time, panting index (number of times the animal breathes through its mouth per unit time), etc. Equipment operating condition feature extraction: number and frequency of fans in operation, spray working cycle, current total energy consumption, etc.; Performance-related feature extraction: the difference between the predicted and baseline values ​​of dairy cow performance, real-time values ​​of performance-influencing risk factors, and the rate of change of risk factors, etc. Based on the physiological stage labels of the dairy cows in the barn, the basic comfort threshold vector corresponding to the physiological stage labels is retrieved. At the same time, the critical warning values ​​of environmental parameters calibrated by the environment-dairy cow performance correlation model are retrieved. These are fused to construct a performance-oriented current environmental state vector. The current environmental state vector is then compared one by one with the basic comfort threshold vector. If any environmental feature (such as dry-bulb temperature) in the current environmental state vector exceeds the threshold in the corresponding basic comfort threshold vector, or if any performance-related feature triggers a risk warning, then a control command is triggered. The scene matching and control module responds to control commands and, in conjunction with sliding window calculations, generates the fused state vector Vt at the current moment. Based on the fused state vector Vt analysis and combined with a preset scene-control mapping rule base, it outputs the environmental control strategy and expected performance recovery target, and generates executable control commands, specifically including: S1: When generating control commands, based on environmental characteristics, cattle status characteristics, and equipment operating condition characteristics, and combined with sliding window calculations, the current fusion state vector Vt = [Et, Ct, Mt, Pt] is generated, where t is greater than zero and t represents the sampling time; Where Et is an environmental feature sub-vector, which includes at least temperature, humidity, THI, ammonia concentration, carbon dioxide concentration, dust concentration, wind speed, and light intensity. Ct is a subvector of cattle state features, which includes at least the average body temperature, average activity level, percentage of lying down time, average rumination time, and a group panting index based on image recognition of the dairy cattle population in the current area. Mt is a feature vector of equipment operating conditions, which includes at least the number of environmental control devices turned on in the current area, the average operating frequency, the operating status, and the real-time energy consumption. Pt is a performance-related feature sub-vector, which includes at least the predicted value of the average milk production performance of the herd, the predicted value of the average growth performance of the herd, the number of performance-influencing risk factors, and the real-time values ​​of the core risk factors. S2: Preprocess the constructed fusion state vector Vt (such as normalization of numerical features, encoding of state features, etc.) and extract temporal extended features (such as the rate of change, acceleration and fluctuation variance of each feature in the past 10 or 30 minutes) to form the extended feature vector KVt; For example: numerical feature normalization: performing Min-Max normalization or Z-score standardization on environmental features (temperature, humidity, THI, ammonia concentration, etc.) and cattle features (body temperature, activity level, panting index, etc.); state feature encoding: encoding binary features such as equipment start / stop status and roller shutter open / close status with 0 / 1; S3: Retrieve the pre-built scene template library and obtain the scene template Sk for each scene in the library. The scene template Sk contains the following elements: Scene identifier: unique ID and semantic name (e.g., "S001-mild heat stress"); Feature pattern vector: Pk = [p1, p2, ..., pn], representing the expected value or expected range of each feature in this scenario. For interval features, it is stored as [min, max]; for precise value features, it is stored as the target value. Feature weight vector: Wk = [w1, w2, ..., wn], representing the importance of each feature in identifying the scenario. The sum of all weights is 1. For example, for heat stress scenarios, the wheezing index and THI have higher weights; for ammonia exceeding the standard scenario, the weight of ammonia concentration is 1, and the weights of other features are 0. Matching threshold: θk, representing the minimum confidence level required to determine if it is the scene in question; Scene Priority: Pyk, used for decision arbitration when multiple scenes are matched successfully at the same time. Scenes with higher priority are responded to first (e.g., "severe heat stress" has higher priority than "mild heat stress"). Related decision recommendations: After a successful match, the recommended control strategy or instruction set index is provided. The control strategy adds a performance recovery orientation, which means that after control, the performance prediction value should return to the excellent range. S4: Calculate the matching degree between the current expanded feature vector KVt and each scene template Sk one by one: For numerical features, the weighted Gaussian similarity function is used to calculate feature similarity; For interval-type features, the degree to which the current value falls within the expected interval is used as the feature similarity. Interval-type feature = max(0, 1 - (distance from the interval boundary / interval width)). If vi is within the interval, the similarity is 1, and the further away it is, the lower the similarity. For state-type features, the feature similarity is calculated using an exact matching method. For example, if vi equals pi, the similarity is 1; otherwise, it is 0. The similarity of each feature is weighted and fused according to the feature weight vector Wk to obtain the comprehensive scene confidence Zxk, that is, the comprehensive scene confidence Zxk = ∑ (wi × Xi), where Xi represents the similarity of the i-th feature, i = 1, 2, 3, ..., n, and n is a natural number greater than zero; S5: Filter out all scene templates with a comprehensive scene confidence score Zxk≥θk as candidate matching scenes; If there is only one candidate scene, it is directly determined as the current single scene label; If there are multiple candidate scenarios, the arbitration logic will be executed: Priority takes precedence: the scenario with the highest priority in Pyk is taken as the main scenario. For example, if "severe heat stress" (priority 10) and "mild heat stress" (priority 5) are matched successfully at the same time, it is determined to be severe heat stress. Confidence-weighted fusion: If multiple high-priority scenarios coexist (such as "peak feeding period" and "heat stress" occurring simultaneously), scenario fusion is performed to generate composite scenario labels, such as "peak feeding period + mild heat stress"; S6: Based on a preset scene-control mapping rule base, match the environmental control strategy and expected performance recovery target corresponding to the scene label, and generate executable control instructions; Environmental control strategies include equipment identification, action type, and operating parameters; The visual dashboard module displays the environmental control strategies and expected performance recovery targets corresponding to the scene labels on the display dashboard.

[0021] Example 3: After the control command corresponding to the environmental control strategy is issued to the environmental control equipment, the dynamic emergency management module activates the real-time data acquisition and feedback mechanism, which specifically includes: The environmental perception data module continuously collects dynamic data on the characteristics of the cowshed environment, the condition of the cows, and the operating conditions of the equipment. At the same time, it collects real-time data related to the performance of the cows (such as real-time milk production and real-time body condition monitoring data). It generates a real-time feature vector ZVt after execution and compares the difference with the fused state vector Vt before the instruction is executed to obtain the feature change difference ΔV. A threshold range [ΔVmin, ΔVmax] is set to determine the effect of regulation. This range is based on the baseline comfort threshold of the dairy cow's physiological stage label and represents the reasonable range of changes in the environment and the cow's condition towards comfort. If ΔV falls within the judgment threshold range and the change value of the performance-related characteristics reaches the expected performance recovery target, it is determined that the control is effective, the current control strategy and equipment operating parameters remain unchanged, and the regular monitoring cycle begins. If ΔV < ΔVmin, or the change value of the performance-related characteristics does not reach the expected performance recovery target, it is determined that the control intensity is insufficient (such as insufficient number of fans to be turned on during heat stress, or the temperature drop does not reach the expected level). Immediately retrieve the enhanced control strategy under the same scenario in the scenario-control mapping rule base, such as increasing the number of fans to be turned on, shortening the spray working interval, increasing the operating power of the wet curtain, etc., and regenerate and issue control commands. If ΔV > ΔVmax, it is determined to be an over-regulation (such as the temperature dropping too quickly, causing cold stress in dairy cows, or the spraying too frequently, causing excessive humidity in the cowshed), triggering a strategy downgrade, such as reducing the number of environmental control devices turned on, extending the equipment operation interval, reducing the equipment operation frequency, and at the same time correcting the equipment operation parameters to a reasonable range. If an abnormal operating condition is detected when the equipment executes commands (such as abnormal current after the fan starts or poor water flow from the spray system), the current control command should be immediately suspended, and equipment fault warning and replacement analysis should be triggered, specifically including: Obtain the unique identifier and installation area of ​​the faulty equipment, retrieve the equipment feature database, and obtain the core control capabilities of the faulty equipment (e.g., if the faulty equipment is a fan with an air volume of 8000 m³ / h). 3 / h, cooling coverage area of ​​50㎡, rated wind speed of 3m / s, etc., and at the same time obtain the control capacity gap in the area after the failure of the faulty equipment (such as an air volume gap of 8000m). 3 / h, wind speed gap 0.5m / s); According to the set screening rules, the alternative equipment is matched. Based on the environmental parameters of the environmental control strategy and the quantitative prediction results of dairy cow performance, the operating parameters of the alternative equipment are adjusted to ensure that the environmental parameters of the area can meet the requirements for dairy cow performance recovery after the alternative equipment is put into operation. The visual dashboard module is used to mark replacement equipment and faulty equipment. Faulty equipment is marked in red and replacement equipment is marked in green. Replacement equipment and faulty equipment are connected by a bidirectional arrow. For example, the screening rules are: equivalent control capability first > energy efficiency first > proximity of installation location first; The first step is to match the control capabilities: select fans whose air volume, air speed, and coverage area are ≤10% different from the faulty fans to ensure that the replacement can fill the gap in control capabilities. Step 2: Energy efficiency matching: If there are multiple equivalent fans, select the fan with the lowest energy consumption-efficiency ratio (i.e., lower energy consumption for the same air volume). Step 3: Matching installation locations: If there are still multiple fans, select a fan whose installation location is adjacent to the faulty fan to ensure the uniformity of wind speed / ventilation in the area after replacement and avoid environmental "dead zones". In addition to marking and displaying faulty and replacement equipment, the visual dashboard module adds information such as environmental parameters, cattle status, and comparison charts of changes in dairy cow performance before and after the execution of control commands, realizing a visualized and integrated display of farm environmental management and dairy cow production performance management.

[0022] Example 4: This invention also proposes a method for managing the breeding environment in smart dairy farms, comprising the following steps: Step 1: Multi-source heterogeneous data acquisition and standardized preprocessing: Environmental parameters, regional image data, physiological parameters of dairy cows, and equipment status parameters of cattle sheds are collected to construct multi-source heterogeneous data. The multi-source heterogeneous data is preprocessed to form standardized multi-dimensional raw data. Step 2: Construct models based on physiological stages and identify performance risk factors: A regression correlation model between environment and dairy cow performance at multiple physiological stages was constructed. The quantitative prediction results of output environmental parameters and dairy cow performance were evaluated, and the risk factors affecting performance were output. Step 3: Feature extraction and scene discrimination trigger control commands: Four core features are extracted from standardized data: environment, cattle status, equipment condition, and performance correlation. The basic comfort threshold vector of the dairy cow herd's physiological stages and the critical warning values ​​of environmental parameters calibrated by the model are combined to construct a performance-oriented current environmental state vector. This vector is compared with the basic comfort threshold vector one by one. If any environmental feature exceeds the threshold or any performance correlation feature triggers a risk warning, a control command is immediately output. Step 4: Integrate state analysis and scenario matching to generate control strategies: By combining sliding window calculation, the fused state vector Vt at the current moment is generated. Based on the analysis of the fused state vector Vt and combined with the preset scene-control mapping rule library, the environmental control strategy and expected performance recovery target are output, and executable control instructions are generated. Step 5: Dynamic emergency feedback and control strategy optimization: After the control command corresponding to the environmental control strategy is issued to the environmental control equipment, the real-time data acquisition and feedback mechanism is activated. Combined with the set control effect judgment threshold range [ΔVmin, ΔVmax], the control effect is judged in three cases and the strategy is optimized. Step Six: Full-process visualization and monitoring; In summary, by collecting and standardizing multi-source data across all dimensions, a solid foundation for management data is established. This not only solves the problems of fragmented and inconsistent formats in traditional livestock data collection, but also provides high-quality, reusable datasets for subsequent model training and scene recognition. Furthermore, based on the growth and milk production characteristics of dairy cows at different physiological stages, a hierarchical environmental-performance sample dataset and a differentiated regression correlation model are constructed. By combining benchmark performance values ​​with critical warning values ​​for high-quality intervals, the performance of dairy cows under the current environment can be accurately quantified and predicted. This allows for the rapid identification of environmental risk factors that lead to performance impairment, thus addressing the problems of traditional "one-size-fits-all" management that fails to consider the differences in the physiological stages of dairy cows. This makes environmental risk prediction more targeted and helps to avoid the negative impact of environmental factors on dairy cow production performance in advance. By using multi-feature fusion for scene recognition and matching, the current breeding scene in the cowshed (including single and composite scenes) can be accurately determined, and corresponding personalized control strategies can be matched. At the same time, performance recovery is used as the control guide, so that the control instructions are no longer limited to meeting environmental parameters, but focus more on ensuring and restoring the production performance of dairy cows. This solves the problem that traditional management only focuses on environmental indicators and is out of touch with dairy cow performance, thus improving the actual value of environmental control. Furthermore, by collecting feedback data in real time, comparing and analyzing the characteristic changes before and after regulation, and combining the specific threshold ranges for physiological stages to determine the regulation effect, differentiated strategies are implemented for three situations: effective regulation, insufficient intensity, and excessive regulation. At the same time, standardized replacement screening rules are formulated for equipment failures, and equivalent replacement equipment is quickly matched and operating parameters are adjusted to effectively fill the gap in control capabilities and avoid environmental loss of control due to equipment failure. This solves the problems of traditional control commands ending as soon as they are issued, lack of feedback optimization, and delayed response to equipment failures.

[0023] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A livestock farming environment management system applied to smart dairy farms, characterized in that, It includes an aquaculture environment control center, an environmental perception data module, a model and impact management module, a scene recognition module, a scene matching and control module, a dynamic emergency management module, and a visual dashboard module; The environmental perception data module is used to collect environmental parameters, regional image data, physiological parameters of dairy cows, and equipment status parameters of cattle sheds, construct multi-source heterogeneous data, preprocess the multi-source heterogeneous data to form standardized multi-dimensional raw data, and send it to the breeding environment management and control center for storage. The Model and Impact Management module is used to construct environmental-performance sample datasets for each physiological stage, and to construct a regression correlation model of environmental-cow performance for multiple physiological stages. It also judges the quantitative prediction results of output environmental parameters and cow performance, and outputs performance impact risk factors. The scene recognition module is used to extract features and perform scene discrimination analysis on standardized multidimensional raw data. It compares the obtained current environment state vector one by one, and if triggered, it outputs control instructions. The scene matching and control module is used to respond to control commands and, in conjunction with sliding window calculation, generate the fusion state vector Vt at the current moment. Based on the analysis of the fusion state vector Vt and combined with the preset scene-control mapping rule library, it outputs the environmental control strategy and the expected performance recovery target, and generates executable control commands. After the control commands corresponding to the environmental control strategy are issued to the environmental control equipment, the dynamic emergency management module activates the real-time data acquisition and feedback mechanism.

2. The aquaculture environment management system applied to a smart dairy farm according to claim 1, characterized in that, The standardized multidimensional raw data analysis process is as follows: Collect environmental parameters, regional image data, dairy cow physiological parameters, and equipment status parameters of cattle sheds to construct multi-source heterogeneous data; The collected multi-source heterogeneous data is preprocessed to form standardized multidimensional raw data.

3. The aquaculture environment management system applied to a smart dairy farm according to claim 1, characterized in that, The analysis process for the performance impact risk factors is as follows: Standardized multidimensional raw data were stratified and filtered according to the physiological stage labels of dairy cows to construct environmental-performance sample datasets for each physiological stage. Based on the environment-performance sample datasets of each physiological stage, a regression correlation model of environment-dairy cow performance in multiple physiological stages is constructed. Real-time environmental parameters are input into the environment-dairy cow performance regression correlation model corresponding to the physiological stage, and the output is the quantitative prediction result of environmental parameters-dairy cow performance, that is, the predicted value of the overall growth and milk production performance of dairy cows in the current environment; at the same time, combined with the pre-set farm dairy cow breed benchmark performance value and high-quality production performance range, the critical warning value of environmental parameters is calibrated - when a change in a certain environmental parameter causes the performance prediction value to be lower than a% of the benchmark value (a>0), or to exceed the high-quality performance range, the environmental parameter is marked as a performance impact risk factor.

4. The aquaculture environment management system applied to a smart dairy farm according to claim 1, characterized in that, The feature extraction and scene discrimination analysis process is as follows: Based on standardized multidimensional raw data, environmental features, cattle status features, equipment operating condition features, and performance correlation features are extracted. Based on the physiological stage labels of the dairy cows in the barn, the basic comfort threshold vector corresponding to the physiological stage labels is retrieved. At the same time, the critical warning values ​​of environmental parameters calibrated by the environment-dairy cow performance correlation model are retrieved. These are fused to construct a performance-oriented current environmental state vector. The current environmental state vector is then compared one by one with the basic comfort threshold vector. If any environmental feature in the current environmental state vector exceeds the threshold in the corresponding basic comfort threshold vector, or if any performance-related feature triggers a risk warning, then a control command is triggered.

5. The aquaculture environment management system applied to a smart dairy farm according to claim 1, characterized in that, The analysis process of the fused state vector Vt is as follows: S1: When generating control commands, based on environmental characteristics, cattle status characteristics, and equipment operating condition characteristics, and combined with sliding window calculations, the current fusion state vector Vt = [Et, Ct, Mt, Pt] is generated, where t is greater than zero and t represents the sampling time; Where Et is the environmental feature vector, Ct is the cattle state feature vector, Mt is the equipment operating condition feature vector, and Pt is the performance correlation feature vector; S2: Preprocess the constructed fusion state vector Vt and extract temporal extended features to form an extended feature vector KVt.

6. The aquaculture environment management system applied to a smart dairy farm according to claim 5, characterized in that, Also includes: S3: Retrieve the pre-built scene template library and obtain the scene template Sk for each scene in the scene template library; S4: Calculate the matching degree between the current extended feature vector KVt and each scene template Sk one by one; The similarity of each feature is weighted and fused according to the feature weight vector Wk to obtain the comprehensive scene confidence Zxk; S5: Filter out all scene templates with a comprehensive scene confidence score Zxk≥θk as candidate matching scenes; If there is only one candidate scenario, it is directly determined as the current single scenario label; if there are multiple candidate scenarios, arbitration logic is executed. Priority takes precedence: the scene with the highest priority in PyK is selected as the main scene; Confidence-weighted fusion: If multiple high-priority scenarios coexist, a composite scenario label is generated; S6: Based on the preset scene-control mapping rule library, match the environmental control strategy and expected performance recovery target corresponding to the scene label, and generate executable control instructions.

7. The aquaculture environment management system applied to a smart dairy farm according to claim 1, characterized in that, The analysis process of the real-time data acquisition and feedback mechanism is as follows: The environmental perception data module continuously collects dynamic data on the characteristics of the cowshed environment, the state of the dairy cows, and the operating conditions of the equipment. At the same time, it collects real-time data related to the performance of the dairy cows, generates a real-time feature vector ZVt after execution, and compares the difference with the fused state vector Vt before the instruction is executed to obtain the feature change difference ΔV. Set a threshold range [ΔVmin, ΔVmax] for judging the effect of regulation: If ΔV falls into the threshold range and the change value of the performance-related feature reaches the expected performance recovery target, the regulation is judged to be effective; if ΔV < ΔVmin, or the change value of the performance-related feature does not reach the expected performance recovery target, the regulation is judged to be insufficient; if ΔV > ΔVmax, the regulation is judged to be excessive. If an abnormal operating condition is detected when the equipment executes instructions, the current control instructions will be immediately suspended, and equipment fault warning and replacement analysis will be triggered.

8. The aquaculture environment management system applied to a smart dairy farm according to claim 7, characterized in that, The equipment fault early warning and replacement analysis process is as follows: Obtain the unique identifier and installation area of ​​the faulty equipment, retrieve the equipment feature database, obtain the core control capabilities of the faulty equipment, and at the same time, identify the control capability gap in the area after the faulty equipment fails. Match alternative devices according to the set filtering rules.

9. A method for managing the breeding environment in a smart dairy farm, applicable to the breeding environment management system for a smart dairy farm as described in any one of claims 1-8, characterized in that: Step 1: Acquisition and Standardization Preprocessing of Multi-Source Heterogeneous Data; Step 2: Construct models based on physiological stages and identify performance risk factors; Step 3: Feature extraction and scene discrimination trigger control commands; Step 4: Integrate state analysis and scenario matching to generate control strategies; Step 5: Dynamic emergency feedback and optimization of control strategies; Step Six: Full-process visualization and monitoring.

Images