A production state evaluation system and method for livestock farms

Abstract

This invention belongs to the field of intelligent management technology for livestock and poultry farming, and specifically relates to a production status assessment system and method for livestock and poultry farms. The system includes a data acquisition module, a data preprocessing and fusion module, an ecological status indicator system construction module, an intelligent assessment model module, a chart visualization module, and an early warning and push module. The data acquisition module collects various production data from the farm in real time. The data preprocessing and fusion module, connected to the data acquisition module, performs cleaning, standardization, anomaly removal, spatiotemporal alignment, and feature fusion on the collected data. This invention supports dynamic indicator weight adjustment, adapting to different livestock species, different farming models, and different seasons. By quantifying production status scores and the Production Efficiency Index (PEI) in real time, it optimizes feeding curves, reduces biased feeding, and is expected to improve feed conversion rate by 5-15%, increase overall production efficiency by more than 10%, and reduce the risk of ineffective drug use and veterinary drug residues.

Description

Technical Field

[0001] This invention relates to the field of intelligent management technology for livestock and poultry farming, specifically to a production status assessment system and method for livestock and poultry farms. Background Technology

[0002] With the rapid development of large-scale and intensive livestock and poultry farming, real-time and accurate assessment of the production status of farms has become a key link in improving farming efficiency, ensuring animal health, reducing disease risks, and achieving green and efficient production.

[0003] The existing technology has the following defects or problems: 1. Traditional farming relies heavily on manual inspections, single-point environmental sensors, or simple monitoring. Environmental data (temperature, humidity, ammonia, etc.), individual behavioral and health data (feeding, drinking, weight, thermal imaging), and production management data (feeding amount, veterinary drug use) are fragmented, misaligned in time and space, and inconsistent in format, making it difficult to accurately judge the overall condition. 2. Existing systems mostly use fixed indicators or human experience to make judgments. The environmental sensitivity of different livestock (pigs / cattle / poultry), different growth stages (piglets / fattening / peak egg production), and different seasons varies greatly, which can lead to misjudgments or redundant indicators. 3. Traditional methods rely on veterinary inspections or simple thresholds, which make it difficult to capture early trend changes (such as abnormal HRI slopes in 3–7 days), leading to the spread of diseases, feed waste, or decreased production efficiency. 4. Existing monitoring systems mostly consist of simple reports or lack push notifications, making it difficult for managers to obtain key information in a timely manner, especially at night or in remote areas.

[0004] It should be noted that the above content falls within the inventor's technical knowledge and does not necessarily constitute prior art. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a production status assessment system and method for livestock and poultry farms, solving the current problems.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a production status assessment system for livestock and poultry farms, comprising: The data acquisition module is used to collect various production data of the farm in real time. The data preprocessing and fusion module is connected to the data acquisition module and can perform cleaning, standardization, anomaly removal, spatiotemporal alignment and feature fusion on the acquired data. An ecological status indicator system construction module, which dynamically selects and weights multi-dimensional evaluation indicators based on livestock and poultry species, growth stages and seasons; The intelligent assessment model module uses deep learning algorithms to perform comprehensive calculations on the fused data and output production status assessment scores, health risk index, production efficiency index, and early warning level. The chart visualization module is used to generate visualization interfaces such as dashboards, heatmaps, trend curves, and risk heatmaps. The early warning and push module is used to realize multi-level threshold judgment, trend analysis and early warning, and push the information to managers, veterinarians and decision-makers through channels such as SMS, APP, WeChat and Enterprise WeChat.

[0007] In some embodiments, the data acquisition module specifically comprises the following sub-modules: The environmental sensing submodule consists of sensors for temperature, humidity, ammonia concentration, carbon dioxide concentration, dust concentration, light intensity, and wind speed, which can effectively collect various environmental data related to production in livestock and poultry farms. The behavior and health perception submodule monitors the individual status of livestock and poultry in real time and collects data through visible light and infrared thermal imaging cameras, electronic ear tags, acoustic sensors, and a weight platform.

[0008] In some embodiments, the data acquisition module further includes the following sub-modules: The production management submodule specifically comprises an automatic feeding / watering record unit and an animal health and medication management unit, wherein: The automatic feeding / drinking recording unit consists of the following sub-units: The real-time feed volume acquisition subunit of the feed tower / feed line is used to collect the total amount of feed actually fed per batch / day through a weight sensor or flow meter. The individual / pen feed intake statistics subunit is used to associate and accumulate the actual feed intake of each animal / pen through electronic feeding stations and electronic ear tag identification technology; The drinking water monitoring subunit is used to collect drinking water volume in real time through a water meter flow sensor and a drinking water tank water level sensor. The feeding curve execution and deviation alarm subunit is used to preset the feeding curve according to the animal's age and weight stage, and to issue an alarm when the actual feed intake deviates from the target curve by more than a set threshold. The animal health and drug management unit is used to update inventory levels in real time, and can also correspond veterinary drug / vaccine usage records with animal electronic ear tags / batch numbers to form a traceability chain.

[0009] In some embodiments, the ecological state indicator system construction module specifically consists of the following sub-modules: The sub-module for selecting adaptation indicators of livestock and poultry species and growth stages has a built-in classification library of livestock and poultry species, which is used to pre-store characteristic parameters of different livestock and poultry such as pigs, cattle and poultry, including manure production coefficient, feed conversion rate benchmark and environmental sensitivity threshold. The growth stage division unit can automatically match the corresponding set of evaluation indicators based on the age, weight stage and physiological period of livestock and poultry.

[0010] In some embodiments, the ecological status indicator system construction module has a built-in dynamic indicator screening logic, which is used to automatically select and activate 10-25 core evaluation indicators from a preset indicator pool based on the livestock and poultry species and current growth stage, avoiding indicator redundancy. The specific algorithm flow is as follows: 1) Obtain the current livestock and poultry species code and current growth stage label; 2) Load the preset livestock-stage-indicator association knowledge base, which includes the recommended indicator set, the required indicator set, the alternative indicator set and the indicator ecological sensitivity coefficient (range 0-1) for each type of livestock and poultry × each growth stage. 3) Complete the first layer of filtering, which specifically involves mandatory matching from category to stage and activation of required indicators, including: The essential core indicator set for the livestock and poultry species and growth stage is directly extracted from the knowledge base. These indicators are directly marked as active and enter the final indicator list, without participating in subsequent screening and elimination. 4) Complete the second layer of filtering, which specifically involves sensitivity priority ranking and preliminary Top-K selection, including: For all remaining indicators, calculate their comprehensive priority score under the current livestock / poultry species and growth stage, using the following formula: ; in: Sensitivity_Score: The pre-set sensitivity of this indicator to the ecological and environmental impact of this stage (0-1, calculated by expert scoring or historical data using the entropy weight method). Historical anomaly detection rate: This indicator is the proportion of batches that exceed the warning threshold in the past 12 months in this or similar batches, reflecting the actual volatility and warning value; Correlation coefficient: Based on Pearson and Spearman correlation analysis, this is the absolute value of the correlation coefficient between the index and the historical comprehensive ecological state index; The weights w1, w2, and w3 are usually set to 0.5, 0.3, and 0.2 by default, and can be adjusted quarterly through AHP or by experts. All candidate metrics are sorted in descending order of Priority Score, and the top N metrics are initially selected, with a target activation number × 1.5-2.0. 5) Complete the third layer of filtering, which specifically involves removing redundant related information, including: For the initially selected index set in the second layer, calculate its pairwise Pearson correlation coefficient matrix between the standards; Hierarchical clustering is used to retain only the item with the highest Priority Score in each highly correlated index cluster, and the rest are marked as redundant and to be removed. 6) Complete the fourth layer of filtering, which specifically involves quantity constraints and coverage verification, including: From the list of redundancies removed, continue to add indicators in descending order of Priority Score until the number of activated indicators reaches the lower limit of the target range. The formula for calculating the multi-dimensional coverage of the current set of activated indicators on the ecological state is as follows: Coverage = (Number of covered dimensions / Total number of dimensions) × 100% + Weighted sum of indicator saturation for each dimension; If the coverage is less than 85% and the number of fecal waste resource utilization indicators is less than 2, the highest priority indicators will be added from the candidate pool until the minimum coverage requirement is met. The final number will be controlled between 10 and 25. If the upper limit is exceeded, the indicators will continue to be removed from the priority score from low to high until the upper limit is reached.

[0011] In some embodiments, the operation process of the intelligent evaluation model module is as follows: 1) First, construct the fused feature matrix, with the following formula: X∈R^{n×t×m}; Where n is the number of samples; t is the length of the time window; and m represents the various features. 2) The overall model architecture formula is as follows: Its embedding layer sinusoidal position encoding formula is: ; Where pos is the time step and i is the dimension index. Hide dimensions in the model; The above algorithm is repeated 4-8 times, with each layer including multi-head attention and a feedforward network; The core formula for multi-head self-attention is: ; Its bullish version formula is: ; ; Where h is the number of heads, and d_k = d_model / h (dimension of each head), this part allows the intelligent assessment model to simultaneously focus on cross-time non-local strong correlations such as ammonia exceeding the standard 3 days ago, yesterday's temperature difference, and today's decrease in water intake; The formula for the pre-network is as follows: .

[0012] In some embodiments, the early warning and push module is specifically composed of the following modules: The real-time data receiving and buffering layer can predict results in real time, with the time interval controlled at 5-60 minutes, and can also maintain the sliding time window buffer of each breeding unit. A multi-level threshold judgment and rule engine layer, wherein the basic threshold classification in the multi-level threshold judgment and rule engine layer can be divided into four levels; The multi-level threshold judgment and rule engine layer has a built-in trend analysis submodule, in which: The formula for calculating the rate of change and slope over 3-7 days is: ΔHRI = HRI(t)-avg(HRI(t-3:t-1)); A multi-channel push execution layer, wherein the core channels of the multi-channel push execution layer include: The mobile app push notification supports rich text, redirection to the corresponding detailed page, voice broadcast, and task confirmation. Enterprise WeChat / WeChat Service Account / WeChat Mini Program, which includes graphic messages, card-based alerts, quick confirmation / feedback buttons, and one-click call to a veterinarian; SMS, which can cover offline scenarios, includes concise text, urgency level indication, key indicator summary and processing link; The alarm closed-loop tracking and feedback layer can record each generated alarm, including the generation time, triggering rule, push channel, recipient, reading / confirmation time, and processing result.

[0013] Another technical problem to be solved by this invention is to provide a method for assessing the production status of livestock and poultry farms, comprising the following steps: Step 1: Through the data acquisition module, using the environmental perception submodule, behavior and health perception submodule, and production management submodule, continuously collect multi-source heterogeneous data from the farm, including environmental data, individual / group behavior and health data, and production management data; Step 2: The collected raw data is preprocessed and multi-source data is fused using the data preprocessing and fusion module. Step 3: Through the ecological status index system construction module, dynamically construct the ecological status index system for the current batch / growth stage, and obtain the basic information of the current assessment object: livestock and poultry species code, current growth stage label (age range or weight stage), and seasonal factors; Step 4: Input the fused feature vector and activation index weights from the above steps into the deep learning model, and output the multi-task prediction results through the model. Step 5: Generate a real-time dashboard: display the current overall score, health risk index, productivity index, and warning level, and plot trend curves based on the score, HRI, and PEI changes over the past 7 / 30 / 90 days; Step Six: Determine the recipients and urgency level of the notification based on the warning level; Step 7: Store the scores, indices, activated indicator sets, warning levels, push records, and subsequent processing results of this evaluation into the database to form a traceable historical evaluation chain, which can be used for subsequent model iteration optimization, indicator library updates, and rule library tuning.

[0014] Compared with the prior art, the present invention provides a production status assessment system and method for livestock and poultry farms, which has the following beneficial effects: This is a production status assessment system and method for livestock and poultry farms. It supports dynamic adjustment of indicator weights, adapts to different livestock species, different breeding modes and different seasons. By quantifying production status scores and production efficiency index (PEI) in real time, it optimizes feeding curves, reduces biased feeding, and is expected to improve feed conversion rate by 5-15%, increase overall production efficiency by more than 10%, and reduce the risk of ineffective drug use and veterinary drug residues. Compared to existing assessment systems, the indicator system in this invention automatically adapts and filters according to type, stage, and season, avoiding a "one-size-fits-all" approach. The assessment is more scientific and targeted, and applicable to pig / cattle / poultry farms of different sizes. At the same time, it adopts multi-channel hierarchical intelligent push and a visual dashboard, enabling managers to shift from "inspection dependence" to "data-driven + anomaly priority", reducing the frequency of on-site inspections by 40-70%, which is especially suitable for large-scale group / remote farms. This system constructs a complete production status assessment system that is data-driven, dynamically intelligent, and closed-loop traceable. It has significant technological advancement, economic practicality, and promotion and application value. It enables multi-dimensional, real-time, and quantitative assessment of the production status of farms, making up for the shortcomings of traditional manual inspections. At the same time, it helps to reduce the incidence of diseases, improve the survival rate and feed conversion efficiency, and achieve cost reduction, efficiency improvement, and green farming. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the production status assessment system for livestock and poultry farms according to the present invention. Detailed Implementation

[0016] 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.

[0017] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0018] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0019] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0020] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0021] Please see Figure 1 In this implementation plan: a production status assessment system for livestock and poultry farms, comprising: The data acquisition module is used to collect various production data from the farm in real time. The data acquisition module consists of the following sub-modules: The environmental sensing submodule consists of sensors for temperature, humidity, ammonia concentration, carbon dioxide concentration, dust concentration, light intensity, and wind speed. It can effectively collect various environmental data related to production in livestock and poultry farms. The Behavior and Health Sensing Submodule monitors the individual status of livestock and poultry in real time and collects data through visible light and infrared thermal imaging cameras, electronic ear tags, acoustic sensors, and a weight platform.

[0022] The production management submodule consists of an automatic feeding / watering record unit and an animal health and medication management unit, wherein: The automatic feeding / drinking recording unit consists of the following sub-units: The real-time feed volume acquisition subunit of the feed tower / feed line is used to collect the total amount of feed actually fed per batch / day through a weight sensor or flow meter. The individual / pen feed intake statistics subunit is used to associate and accumulate the actual feed intake of each animal / pen through electronic feeding stations and electronic ear tag identification technology; The drinking water monitoring subunit is used to collect drinking water volume in real time through a water meter flow sensor and a drinking water tank water level sensor. The feeding curve execution and deviation alarm subunit is used to preset the feeding curve according to the animal's age and weight stage, and to issue an alarm when the actual feed intake deviates from the target curve by more than a set threshold. The animal health and drug management unit is used to update inventory levels in real time, and can also match veterinary drug / vaccine usage records with animal electronic ear tags / batch numbers to form a traceability chain; The data preprocessing and fusion module is connected to the data acquisition module. It can clean, standardize, remove anomalies, align the data in time and space, and fuse features. The ecological status indicator system construction module selects and weights multi-dimensional evaluation indicators based on livestock and poultry species, growth stages, and seasonal dynamics. The ecological status indicator system construction module consists of the following sub-modules: The submodule for selecting adaptation indicators for livestock and poultry species and growth stages has a built-in classification library of livestock and poultry species, which is used to pre-store characteristic parameters of different livestock and poultry such as pigs, cattle and poultry, including manure production coefficient, feed conversion rate benchmark and environmental sensitivity threshold. The growth stage division unit can automatically match the corresponding set of evaluation indicators based on the age, weight stage and physiological period of livestock and poultry.

[0023] The ecological status indicator system construction module has a built-in dynamic indicator selection logic, which automatically selects and activates 10-25 core assessment indicators from a preset indicator pool based on the livestock and poultry species and their current growth stage, avoiding indicator redundancy. The specific algorithm flow is as follows: 1) Obtain the current livestock and poultry species code and current growth stage label; 2) Load the preset livestock-stage-indicator association knowledge base, which includes the recommended indicator set, the required indicator set, the alternative indicator set and the indicator ecological sensitivity coefficient (range 0-1) for each type of livestock and poultry × each growth stage. 3) Complete the first layer of filtering, which specifically involves mandatory matching from category to stage and activation of required indicators, including: The essential core indicator set for the livestock and poultry species and growth stage is directly extracted from the knowledge base. These indicators are directly marked as active and enter the final indicator list, without participating in subsequent screening and elimination. 4) Complete the second layer of filtering, which specifically involves sensitivity priority ranking and preliminary Top-K selection, including: For all remaining indicators, calculate their comprehensive priority score under the current livestock / poultry species and growth stage, using the following formula: ; in: Sensitivity_Score: The pre-set sensitivity of this indicator to the ecological and environmental impact of this stage (0-1, calculated by expert scoring or historical data using the entropy weight method). Historical anomaly detection rate: This indicator is the proportion of batches that exceed the warning threshold in the past 12 months in this or similar batches, reflecting the actual volatility and warning value; Correlation coefficient: Based on Pearson and Spearman correlation analysis, this is the absolute value of the correlation coefficient between the index and the historical comprehensive ecological state index; The weights w1, w2, and w3 are usually set to 0.5, 0.3, and 0.2 by default, and can be adjusted quarterly through AHP or by experts. All candidate metrics are sorted in descending order of Priority Score, and the top N metrics are initially selected, with a target activation number × 1.5-2.0. 5) Complete the third layer of filtering, which specifically involves removing redundant related information, including: For the initially selected index set in the second layer, calculate its pairwise Pearson correlation coefficient matrix between the standards; Hierarchical clustering is used to retain only the item with the highest Priority Score in each highly correlated index cluster, and the rest are marked as redundant and to be removed. 6) Complete the fourth layer of filtering, which specifically involves quantity constraints and coverage verification, including: From the list of redundancies removed, continue to add indicators in descending order of Priority Score until the number of activated indicators reaches the lower limit of the target range. The formula for calculating the multi-dimensional coverage of the current set of activated indicators on the ecological state is as follows: Coverage = (Number of covered dimensions / Total number of dimensions) × 100% + Weighted sum of indicator saturation for each dimension; When the coverage is less than 85% and the number of fecal waste resource utilization indicators is less than 2, the highest priority indicators will be added from the candidate pool until the minimum coverage requirement is met. The final number will be controlled between 10 and 25. If the upper limit is exceeded, the indicators will continue to be removed from the priority score from low to high until the upper limit is reached. The intelligent assessment model module uses deep learning algorithms to perform comprehensive calculations on the fused data and output production status assessment scores, health risk index, production efficiency index, and early warning level. The specific operation process of the intelligent evaluation model module is as follows: 1) First, construct the fused feature matrix, with the following formula: X∈R^{n×t×m}; Where n is the number of samples; t is the length of the time window; and m represents the various features. 2) The overall model architecture formula is as follows: Its embedding layer sinusoidal position encoding formula is: ; Where pos is the time step and i is the dimension index. Hide dimensions in the model; The above algorithm is repeated 4-8 times, with each layer including multi-head attention and a feedforward network; The core formula for multi-head self-attention is: ; Its bullish version formula is: ; ; Where h is the number of heads, and d_k = d_model / h (dimension of each head), this part allows the intelligent assessment model to simultaneously focus on cross-time non-local strong correlations such as ammonia exceeding the standard 3 days ago, yesterday's temperature difference, and today's decrease in water intake; The formula for the pre-network is as follows: .

[0024] The chart visualization module is used to generate visualization interfaces such as dashboards, heatmaps, trend curves, and risk heatmaps. The early warning and push module is used to realize multi-level threshold judgment, trend analysis and early warning, and push the information to managers, veterinarians and decision-makers through channels such as SMS, APP, WeChat and WeChat Work. The early warning and push module consists of the following modules: The real-time data receiving and buffering layer can predict results in real time, with the time interval controlled at 5-60 minutes, and can also maintain the sliding time window buffer of each breeding unit. The multi-level threshold judgment and rule engine layer can be divided into four levels based on the basic threshold classification. The corresponding urgency level and receiving role for Level 4 alarms are shown in the table below:

[0025] The multi-level threshold judgment and rule engine layer has a built-in trend analysis submodule, in which: The formula for calculating the rate of change and slope over 3-7 days is: ΔHRI = HRI(t)-avg(HRI(t-3:t-1)); The multi-channel push execution layer includes the following core channels: The mobile app push notification supports rich text, redirection to the corresponding detailed page, voice broadcast, and task confirmation. Enterprise WeChat / WeChat Service Account / WeChat Mini Program, which includes graphic messages, card-based alerts, quick confirmation / feedback buttons, and one-click call to a veterinarian; SMS, which can cover offline scenarios, includes concise text, urgency level indication, key indicator summary and processing link; The alarm closed-loop tracking and feedback layer can record every alarm generated, including the generation time, triggering rule, push channel, recipient, reading / confirmation time, and processing result.

[0026] Based on the aforementioned production status assessment system for livestock and poultry farms, a method for assessing the production status of livestock and poultry farms is proposed, comprising the following steps: Step 1: Through the data acquisition module, using the environmental perception submodule, behavior and health perception submodule, and production management submodule, continuously collect multi-source heterogeneous data from the farm, including environmental data, individual / group behavior and health data, and production management data; Step 2: The collected raw data is preprocessed and multi-source data is fused using the data preprocessing and fusion module. Step 3: Through the ecological status index system construction module, dynamically construct the ecological status index system for the current batch / growth stage, and obtain the basic information of the current assessment object: livestock and poultry species code, current growth stage label (age range or weight stage), and seasonal factors; Step 4: Input the fused feature vector and activation index weights from the above steps into the deep learning model, and output the multi-task prediction results through the model. Step 5: Generate a real-time dashboard: display the current overall score, health risk index, productivity index, and warning level, and plot trend curves based on the score, HRI, and PEI changes over the past 7 / 30 / 90 days; Step Six: Determine the recipients and urgency level of the notification based on the warning level; Step 7: Store the scores, indices, activated indicator sets, warning levels, push records, and subsequent processing results of this evaluation into the database to form a traceable historical evaluation chain, which can be used for subsequent model iteration optimization, indicator library updates, and rule library tuning.

[0027] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0028] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A production status evaluation system for livestock farms, characterized by, include: The data acquisition module is used to collect various production data of the farm in real time. The data preprocessing and fusion module is connected to the data acquisition module and can perform cleaning, standardization, anomaly removal, spatiotemporal alignment and feature fusion on the acquired data. An ecological status indicator system construction module, which dynamically selects and weights multi-dimensional evaluation indicators based on livestock and poultry species, growth stages and seasons; The intelligent assessment model module uses deep learning algorithms to perform comprehensive calculations on the fused data and output production status assessment scores, health risk index, production efficiency index, and early warning level. The chart visualization module is used to generate visualization interfaces such as dashboards, heatmaps, trend curves, and risk heatmaps. The early warning and push module is used to realize multi-level threshold judgment, trend analysis and early warning, and push the information to managers, veterinarians and decision-makers through channels such as SMS, APP, WeChat and Enterprise WeChat.

2. The production status evaluation system for livestock and poultry farms according to claim 1, characterized in that, The data acquisition module is specifically composed of the following sub-modules: The environmental sensing submodule consists of sensors for temperature, humidity, ammonia concentration, carbon dioxide concentration, dust concentration, light intensity, and wind speed, which can effectively collect various environmental data related to production in livestock and poultry farms. The behavior and health perception submodule monitors the individual status of livestock and poultry in real time and collects data through visible light and infrared thermal imaging cameras, electronic ear tags, acoustic sensors, and a weight platform.

3. The production status evaluation system for livestock and poultry farms according to claim 2, characterized in that, The data acquisition module also includes the following sub-modules: The production management submodule specifically comprises an automatic feeding / watering record unit and an animal health and medication management unit, wherein: The automatic feeding / drinking recording unit consists of the following sub-units: The real-time feed volume acquisition subunit of the feed tower / feed line is used to collect the total amount of feed actually fed per batch / day through a weight sensor or flow meter. The individual / pen feed intake statistics subunit is used to associate and accumulate the actual feed intake of each animal / pen through electronic feeding stations and electronic ear tag identification technology; The drinking water monitoring subunit is used to collect drinking water volume in real time through a water meter flow sensor and a drinking water tank water level sensor. The feeding curve execution and deviation alarm subunit is used to preset the feeding curve according to the animal's age and weight stage, and to issue an alarm when the actual feed intake deviates from the target curve by more than a set threshold. The animal health and drug management unit is used to update inventory levels in real time, and can also correspond veterinary drug / vaccine usage records with animal electronic ear tags / batch numbers to form a traceability chain.

4. The production status assessment system for livestock and poultry farms according to claim 1, characterized in that, The ecological status indicator system construction module is specifically composed of the following sub-modules: The sub-module for selecting adaptation indicators of livestock and poultry species and growth stages has a built-in classification library of livestock and poultry species, which is used to pre-store characteristic parameters of different livestock and poultry such as pigs, cattle and poultry, including manure production coefficient, feed conversion rate benchmark and environmental sensitivity threshold. The growth stage division unit can automatically match the corresponding set of evaluation indicators based on the age, weight stage and physiological period of livestock and poultry.

5. The production status assessment system for livestock and poultry farms according to claim 4, characterized in that, The ecological status indicator system construction module has a built-in dynamic indicator screening logic, which is used to automatically select and activate 10-25 core evaluation indicators from a preset indicator pool based on the livestock and poultry species and current growth stage, avoiding indicator redundancy. The specific algorithm flow is as follows: 1) Obtain the current livestock and poultry species code and current growth stage label; 2) Load the preset livestock-stage-indicator association knowledge base, which includes the recommended indicator set, the required indicator set, the alternative indicator set and the indicator ecological sensitivity coefficient (range 0-1) for each type of livestock and poultry × each growth stage. 3) Complete the first layer of filtering, which specifically involves mandatory matching from category to stage and activation of required indicators, including: The essential core indicator set for the livestock and poultry species and growth stage is directly extracted from the knowledge base. These indicators are directly marked as active and enter the final indicator list, without participating in subsequent screening and elimination. 4) Complete the second layer of filtering, which specifically involves sensitivity priority ranking and preliminary Top-K selection, including: For all remaining indicators, calculate their comprehensive priority score under the current livestock / poultry species and growth stage, using the following formula: ； in: Sensitivity_Score: The pre-set sensitivity of this indicator to the ecological and environmental impact of this stage (0-1, calculated by expert scoring or historical data using the entropy weight method). Historical anomaly detection rate: This indicator is the proportion of batches that exceed the warning threshold in the past 12 months in this or similar batches, reflecting the actual volatility and warning value; Correlation coefficient: Based on Pearson and Spearman correlation analysis, this is the absolute value of the correlation coefficient between the index and the historical comprehensive ecological state index; The weights w1, w2, and w3 are usually set to 0.5, 0.3, and 0.2 by default, and can be adjusted quarterly through AHP or by experts. All candidate metrics are sorted in descending order of Priority Score, and the top N metrics are initially selected, with a target activation number × 1.5-2.

0. 5) Complete the third layer of filtering, which specifically involves removing redundant related information, including: For the initially selected index set in the second layer, calculate its pairwise Pearson correlation coefficient matrix between the standards; Hierarchical clustering is used to retain only the item with the highest Priority Score in each highly correlated index cluster, and the rest are marked as redundant and to be removed. 6) Complete the fourth layer of filtering, which specifically involves quantity constraints and coverage verification, including: From the list of redundancies removed, continue to add indicators in descending order of Priority Score until the number of activated indicators reaches the lower limit of the target range. The formula for calculating the multi-dimensional coverage of the current set of activated indicators on the ecological state is as follows: Coverage = (Number of covered dimensions / Total number of dimensions) × 100% + Weighted sum of indicator saturation for each dimension; If the coverage is less than 85% and the number of fecal waste resource utilization indicators is less than 2, the highest priority indicators will be added from the candidate pool until the minimum coverage requirement is met. The final number will be controlled between 10 and 25. If the upper limit is exceeded, the indicators will continue to be removed from the PriorityScore from low to high until the upper limit is reached.

6. The production status assessment system for livestock and poultry farms according to claim 1, characterized in that, The specific operation process of the intelligent evaluation model module is as follows: 1) First, construct the fused feature matrix, with the following formula: X∈R^{n×t×m}; Where n is the number of samples; t is the length of the time window; and m represents the various features. 2) The overall model architecture formula is as follows: Its embedding layer sinusoidal position encoding formula is: ； Where pos is the time step and i is the dimension index. Hide dimensions in the model; The above algorithm is repeated 4-8 times, with each layer including multi-head attention and a feedforward network; The core formula for multi-head self-attention is: ； Its bullish version formula is: ； ； Where h is the number of heads, and d_k = d_model / h (dimension of each head), this part allows the intelligent assessment model to simultaneously focus on cross-time non-local strong correlations such as ammonia exceeding the standard 3 days ago, yesterday's temperature difference, and today's decrease in water intake; The formula for the pre-network is as follows: 。 7. The production status assessment system for livestock and poultry farms according to claim 1, characterized in that, The early warning and push module is specifically composed of the following modules: The real-time data receiving and buffering layer can predict results in real time, with the time interval controlled at 5-60 minutes, and can also maintain the sliding time window buffer of each breeding unit. A multi-level threshold judgment and rule engine layer, wherein the basic threshold classification in the multi-level threshold judgment and rule engine layer can be divided into four levels; The multi-level threshold judgment and rule engine layer has a built-in trend analysis submodule, in which: The formula for calculating the rate of change and slope over 3-7 days is: ΔHRI = HRI(t)-avg(HRI(t-3:t-1)); A multi-channel push execution layer, wherein the core channels of the multi-channel push execution layer include: The mobile app push notification supports rich text, redirection to the corresponding detailed page, voice broadcast, and task confirmation. Enterprise WeChat / WeChat Service Account / WeChat Mini Program, which includes graphic messages, card-based alerts, quick confirmation / feedback buttons, and one-click call to a veterinarian; SMS, which can cover offline scenarios, includes concise text, urgency level indication, key indicator summary and processing link; The alarm closed-loop tracking and feedback layer can record every alarm generated, including the generation time, triggering rule, push channel, recipient, reading / confirmation time, and processing result.

8. A method for assessing the production status of livestock and poultry farms, characterized in that, Includes the following steps: Step 1: Through the data acquisition module, using the environmental perception submodule, behavior and health perception submodule, and production management submodule, continuously collect multi-source heterogeneous data from the farm, including environmental data, individual / group behavior and health data, and production management data; Step 2: The collected raw data is preprocessed and multi-source data is fused using the data preprocessing and fusion module. Step 3: Through the ecological status index system construction module, dynamically construct the ecological status index system for the current batch / growth stage, and obtain the basic information of the current assessment object: livestock and poultry species code, current growth stage label (age range or weight stage), and seasonal factors; Step 4: Input the fused feature vector and activation index weights from the above steps into the deep learning model, and output the multi-task prediction results through the model. Step 5: Generate a real-time dashboard: display the current overall score, health risk index, productivity index, and warning level, and plot trend curves based on the score, HRI, and PEI changes over the past 7 / 30 / 90 days; Step Six: Determine the recipients and urgency level of the notification based on the warning level; Step 7: Store the scores, indices, activated indicator sets, warning levels, push records, and subsequent processing results of this evaluation into the database to form a traceable historical evaluation chain, which can be used for subsequent model iteration optimization, indicator library updates, and rule library tuning.

Images