A broiler industry full-chain benchmarking decision platform based on big data and intelligent models

Abstract

This invention relates to the field of smart agriculture and livestock information technology, and proposes a benchmarking and decision-making platform for the entire broiler industry chain based on big data and intelligent models. The platform includes a multi-source sensing and acquisition module, a full-chain data platform module, an intelligent model engine module, a full-chain benchmarking and decision-making module, and a multi-terminal interaction module. The multi-source sensing and acquisition module is used to acquire data on the entire broiler industry chain, including production, environment, equipment, supply chain, market, and industry benchmarks. The full-chain data platform module is used for data access, cleaning, standardization, integrated governance, and blockchain notarization to form unified data assets. The intelligent model engine module is used to integrate various machine learning and deep learning models. The full-chain benchmarking and decision-making module is used for horizontal benchmarking, vertical benchmarking, full-chain benchmarking, weakness diagnosis, optimization recommendation, and effect tracking. The multi-terminal interaction module is used for output to PCs, mobile devices, large screens, and API interfaces to achieve real-time monitoring, early warning, and decision execution.

Description

Technical Field

[0001] This invention relates to the field of smart agriculture and livestock information technology, specifically to a benchmarking decision-making platform for the entire broiler industry chain based on big data and intelligent models. Background Technology

[0002] Currently, the broiler industry is undergoing a critical period of transformation from traditional farming methods to a modern production model that is intelligent, refined, and data-driven. With the rapid development of the Internet of Things, big data, artificial intelligence, and blockchain technologies, digital transformation in the agricultural and livestock sectors has become a core path to enhance industry competitiveness. National policies are also continuously promoting the development of smart agriculture and livestock informatization, requiring measures to ensure food safety, achieve full traceability, and improve production efficiency and resource utilization.

[0003] Against this backdrop, broiler farming and processing enterprises face fierce market competition and stringent regulatory requirements. The industry's goals span the entire supply chain, including: precise nutrition and disease control in the farming stage, loss control and quality improvement in the slaughtering and processing stage, cold chain security in logistics and warehousing, and market risk management in operational management. Achieving these goals requires integrating data across the entire "farm-to-table" chain and transforming that data into actionable decisions.

[0004] However, traditional production management methods rely heavily on personal experience and scattered data recording, which has certain technical shortcomings:

[0005] First, data from various stages such as breeding, slaughtering, logistics, and sales are recorded by different equipment and systems, with inconsistent standards and heterogeneous formats. For example, environmental sensor data from farms, vision system data from slaughter lines, GPS temperature and humidity data from logistics vehicles, and ERP data from enterprises are independent of each other and cannot be effectively linked and integrated.

[0006] Second, key production decisions (such as feed formulation, timing of slaughter, and disease assessment) rely heavily on the personal experience of technical personnel, making them highly subjective and difficult to quantify and optimize. The lack of prediction and diagnostic models trained on big data hinders preventative management and maximizes efficiency.

[0007] Third, companies struggle to scientifically assess the gap between their production efficiency and cost control and industry benchmarks or theoretical best practices (i.e., "benchmarking"), making it difficult to accurately pinpoint operational weaknesses. Simultaneously, the lack of transparency across the entire product supply chain hinders reliable food safety traceability, impacting brand value and consumer trust.

[0008] In view of this, the present invention proposes a benchmarking decision-making platform for the entire broiler industry chain based on big data and intelligent models. Summary of the Invention

[0009] To address this, the present invention provides a benchmarking decision-making platform for the entire broiler industry chain based on big data and intelligent models, in order to solve the problem that existing technologies cannot effectively correlate and integrate data from various stages.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] A benchmarking and decision-making platform for the entire broiler industry chain based on big data and intelligent models includes a multi-source sensing and acquisition module, a full-chain data platform module, an intelligent model engine module, a full-chain benchmarking and decision-making module, and a multi-terminal interaction module;

[0012] The multi-source sensing and acquisition module is used to acquire data on the entire broiler chicken production chain, including production, environment, equipment, supply chain, market, and industry benchmarks.

[0013] The full-chain data platform module is used for data access, cleaning, standardization, integrated governance and blockchain notarization to form a unified data asset;

[0014] The intelligent model engine module is used to integrate multiple machine learning and deep learning models;

[0015] The full-chain benchmarking decision module is used for horizontal benchmarking, vertical benchmarking, full-chain benchmarking, weakness diagnosis, optimization recommendation and effect tracking;

[0016] The multi-terminal interaction module is used for output to PCs, mobile devices, large screens, and API interfaces to achieve real-time monitoring, early warning, and decision execution.

[0017] Preferably, the multi-source sensing and acquisition module includes:

[0018] In the breeding process: intelligent chicken scales, temperature and humidity sensors, carbon dioxide sensors, high-definition cameras, RFID tags, and automatic feeding and drinking record equipment;

[0019] Slaughtering and processing: production line visual inspection system, automated cutting line data acquisition points, pre-cooling and quick-freezing temperature monitoring equipment;

[0020] Logistics and warehousing: vehicle-mounted GPS / BeiDou positioning and temperature and humidity recorders, cold storage environment monitoring equipment;

[0021] Management and Market Links: Data interfaces for Enterprise Resource Planning (ERP), Manufacturing Execution System (MES), and Supply Chain Management System (SLM), as well as access interfaces for publicly available market prices, industry standards, and policy and regulatory data sources.

[0022] Preferably, the intelligent model engine module includes:

[0023] The dynamic nutritional requirement model dynamically calculates the optimal nutritional requirements based on parameters such as broiler breed, age, weight, and ambient temperature, and combines them with a raw material price database to generate the lowest cost feed formula.

[0024] The growth and feed conversion ratio prediction model uses historical growth curves, environmental data, and feeding records to predict the date when the flock will reach the target weight, and recommends the most economically efficient slaughter time by taking into account factors such as feed costs, market prices, and pen turnover.

[0025] The disease early warning model is based on video and audio data of chicken behavior, individual body temperature and environmental data to build a multimodal fusion diagnostic model to achieve early warning and intelligent diagnosis of diseases;

[0026] The loss control model predicts carcass quality based on live animal indicators at the end of the breeding period, and uses machine vision to monitor and grade the slaughtering and processing process in real time, optimize the process flow, and reduce losses.

[0027] The whole-chain benchmarking diagnostic model integrates cost data from all stages, conducts input-output analysis, and identifies key points for cost control.

[0028] The risk prediction model integrates historical and real-time market data to predict price trends and assist in sales decisions.

[0029] Preferably, the dynamic nutritional requirement model includes daily metabolizable energy requirement, daily crude protein requirement, and daily lysine requirement;

[0030] The formula for calculating the daily metabolic energy requirement ME(d) is as follows:

[0031] ;

[0032] Where d is the age of the animals, BW(d) is the actual weight measured on day d, ΔT(d) is the difference between the temperature of the enclosure and the suitable temperature, ADG(d) is the average daily weight gain, ηenv(d) is the comprehensive environmental correction coefficient, and a,b,c,k are the breed-specific fitting coefficients.

[0033] The formula for calculating the daily crude protein requirement CP(d) is as follows:

[0034] ;

[0035] Wherein, CPmaint is the crude protein requirement for maintenance, CPgain is the crude protein requirement per unit weight gain, ε(d) is the stress, immunity, or disease correction term, and UCP is the crude protein utilization rate.

[0036] The formula for calculating the daily lysine requirement (Lys(d)) is as follows:

[0037] ;

[0038] Where α, β, and γ are regression coefficients.

[0039] Preferably, the growth and feed conversion ratio prediction model includes a daily cumulative feed conversion ratio function and a multi-factor dynamic FCR prediction function;

[0040] The formula for calculating the cumulative feed conversion ratio function for the day is as follows:

[0041] ;

[0042] Wherein, FI(t) is the actual food intake on day t, BW(d) is the predicted weight on day d, and BW(1) is the weight upon entering the enclosure;

[0043] The calculation formula for the multi-factor dynamic FCR prediction function is as follows:

[0044] ;

[0045] in, For the predicted feed conversion ratio on day d, To predict daily weight gain, The temperature inside the building. For the optimal temperature, This refers to the concentration of ammonia gas. For stocking density, For the sake of egg ratio, These are the model fitting coefficients. For random perturbation and health correction terms.

[0046] Preferably, the disease early warning model includes a core risk probability function, a single-class feature risk quantification function, and a disease classification early warning determination function;

[0047] The formula for the core risk probability function is as follows:

[0048] ;

[0049] in, : The overall risk probability of broiler disease on day d, with a value range of [0,1]. The higher the value, the greater the risk.

[0050] Sigmoid activation function This is used to normalize the weighted result to the 0-1 interval and determine the adaptation probability.

[0051] The weight coefficients of the i-th type of risk feature are fitted by training with big data samples and calibrated with industry benchmark data, and the weights sum to 1.

[0052] The standardized risk feature vector of class i on day d covers five core dimensions:

[0053] Environmental characteristics Temperature deviation, ammonia concentration, humidity exceeding the standard, and abnormal ventilation volume;

[0054] Behavioral characteristics : Rate of decrease in feed intake, fluctuation in water consumption, decrease in activity frequency, and rate of recognition of lethargy;

[0055] Physiological and immune characteristics : Antibody titer failure rate, daily weight gain drop rate, and daily mortality rate increase;

[0056] Management characteristics : Overcrowding, failure to meet disinfection frequency standards, and frequency of stress events;

[0057] External public opinion characteristics Regional disease incidence rate, and risk value of disease transmission during transportation / introduction;

[0058] Model bias term, dynamically calibrated based on broiler breed, breeding cycle, and regional disease history;

[0059] The total number of risk features is n=16 in this model, covering all dimensions of disease triggers;

[0060] The formula for the single-class feature risk quantification function is as follows:

[0061] ;

[0062] in, : Measured value of feature on day d;

[0063] : Optimal threshold for features (suitable value for broiler farming);

[0064] Characteristic danger threshold (the critical value that triggers an epidemic);

[0065] The formula for the disease classification and early warning determination function is as follows:

[0066] ;

[0067] in, : The predicted risk probability of the epidemic on day t in the future (t=1,2,3, to achieve short-term advance warning);

[0068] Long Short-Term Memory (LSTM) neural network fitting function to uncover temporal risk patterns;

[0069] The historical risk data time series length is set by taking the risk data from the past 7 days as the time series input; the remaining parameters are consistent with the basic model to ensure that the data are from the same source and the model is linked.

[0070] Preferably, the function formula of the loss control model is:

[0071] ;

[0072] in, The overall loss rate (%) of the entire chain in time period t, covering slaughtering and deboning loss, cold chain spoilage loss, and storage damage loss;

[0073] : The model's basic bias term is calibrated based on the industry benchmark loss rate;

[0074] : The weight coefficient of the i-th type of loss influencing factor, with a weight sum of 1;

[0075] Xi(t): The feature vector of standardized loss in time period t, with core dimensions including:

[0076] Processing end: slaughter rate X1(t), deboning process proficiency X2(t), carcass temperature X3(t);

[0077] Cold chain end: cold chain temperature and humidity deviation X4 (t), transportation bump coefficient X5 (t), circulation time deviation X6 (t);

[0078] Warehousing: Warehouse stacking density x7 (t), remaining shelf life percentage of products x8 (t);

[0079] The temporal influence coefficient of historical loss rate (0<β<1) reflects the temporal continuity of loss.

[0080] : Random disturbance term, covering sudden factors.

[0081] Preferably, the function formula of the whole-chain benchmarking diagnostic model is: ;

[0082] in, : The overall operational efficiency value of the benchmark entity (such as a certain farm) on day d, with a value of [value missing]. The closer it is to 1, the better the efficiency.

[0083] : The r-th output indicator of the j-th benchmarking unit on day d;

[0084] : The i-th input indicator of the j-th benchmarking unit on day d;

[0085] : Output / input indicator values ​​of the subject being diagnosed;

[0086] The benchmark unit weighting coefficient reflects the reference value of different benchmark units;

[0087] Total number of benchmarking units;

[0088] Number of dimensions for output / input indicators.

[0089] Preferably, the function formula of the risk prediction model is:

[0090] ;

[0091] Where k=7: the time series length of historical risk data;

[0092] The time-series characteristics of four types of risks—disease, cost, price, and loss—over the past 7 days;

[0093] Risk level determination mapping function, t=1,2,3 to achieve short-term early warning, the model data is linked with the basic models of disease, loss, nutrition and other basic models.

[0094] Preferably, the core functions of the full-chain benchmarking decision-making module include:

[0095] Dynamic benchmarking analysis: Users define benchmarking targets, and the platform automatically compares the key performance indicators of this session / batch with the benchmarking targets in multiple dimensions, and displays the gaps intuitively in the form of dashboards, trend charts or radar charts;

[0096] Intelligent decision support: Based on benchmarking results and model analysis, the platform automatically generates optimization suggestion reports, providing specific suggested dates for maximizing benefits when hogs are ready for slaughter.

[0097] Full-chain visual traceability: Utilizing blockchain technology, key data from each stage are stored on the chain for evidence, generating an unalterable traceability QR code;

[0098] Early warning and alarm center: Real-time monitoring of key indicators in all aspects. Once the indicators deviate from the preset threshold or the model prediction range, early warning information will be sent to relevant personnel immediately via APP and SMS.

[0099] Knowledge base and expert system: Integrates industry standards, breeding specifications, disease atlases, and medication guidelines, and combines them with a case library to provide online Q&A and decision support for frontline personnel.

[0100] The present invention has the following advantages:

[0101] 1. By automating the collection of data from each stage through a "multi-source sensing and acquisition module," and then cleaning, standardizing, integrating, and managing the data through a "full-chain data platform module," and storing it on the blockchain, the pain points of inconsistent data standards, heterogeneous formats, and difficulty in data correlation and integration across different stages of the broiler industry are fundamentally resolved. This forms a unified and reliable data asset, providing a high-quality and highly consistent data foundation for subsequent intelligent analysis and decision-making, and utilizing blockchain technology to ensure the authenticity and traceability of the data.

[0102] 2. The full-chain benchmarking decision-making module can transform the model's analysis results into intuitive visualizations (such as dashboards and radar charts) and specific optimization suggestion reports. It supports users in conducting multi-dimensional dynamic benchmarking against custom benchmarks (such as best-in-class and historical best), clearly demonstrating the gaps. Simultaneously, this module integrates an early warning center, which can issue real-time alerts for deviations from thresholds or prediction ranges, driving proactive intervention by managers and forming a closed-loop management system of "monitoring-analysis-decision-execution-optimization".

[0103] 3. By integrating multi-source risk time-series data such as market, disease, cost, and loss through risk prediction models, we can predict the comprehensive risk level in the short term, help enterprises cope with uncertainties such as price fluctuations and disease spread in advance, and improve their operational resilience and risk resistance capabilities.

[0104] 4. By utilizing blockchain technology to immutably store key data across the entire chain, a unique traceability QR code can be generated. Consumers or regulators can scan the code to access information about the entire process of a product from breeding to distribution, greatly enhancing food safety transparency, product credibility, and brand value.

[0105] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. Attached Figure Description

[0106] To more intuitively illustrate the prior art and this application, exemplary drawings are provided below. It should be understood that the specific shapes and structures shown in the drawings should not generally be regarded as limiting conditions for implementing this application; for example, based on the technical concept disclosed in this application and the exemplary drawings, those skilled in the art are able to easily make conventional adjustments or further optimizations to the addition / reduction / classification, specific shapes, positional relationships, connection methods, size ratios, etc. of certain units (components).

[0107] Figure 1 This is a system block diagram of a broiler industry whole-chain benchmarking decision-making platform based on big data and intelligent models according to the present invention;

[0108] Figure 2This is a system block diagram of the multi-source sensing and acquisition module of the present invention;

[0109] Figure 3 This is a system block diagram of the intelligent model engine module of the present invention;

[0110] Figure 4 This is a system block diagram of the full-chain benchmarking decision module of the present invention.

[0111] In the diagram: 10. Multi-source sensing and acquisition module; 20. Full-chain data platform module; 30. Intelligent model engine module; 40. Full-chain benchmarking and decision-making module; 50. Multi-terminal interaction module. Detailed Implementation

[0112] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these embodiments are merely for further explanation of the present invention and should not be construed as limiting the scope of protection of the present invention. Technical engineers in the field can make some non-essential improvements and adjustments to the present invention based on the above-described content. 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.

[0113] Please see Figure 1-4 A benchmarking and decision-making platform for the entire broiler industry chain based on big data and intelligent models, including a multi-source sensing and acquisition module 10, a full-chain data platform module 20, an intelligent model engine module 30, a full-chain benchmarking and decision-making module 40, and a multi-terminal interaction module 50;

[0114] The multi-source sensing and acquisition module 10 is used to acquire data on the entire broiler chicken production chain, including production, environment, equipment, supply chain, market, and industry benchmarks.

[0115] The full-chain data platform module 20 is used for data access, cleaning, standardization, integrated governance and blockchain notarization to form a unified data asset;

[0116] The intelligent model engine module 30 is used to integrate various machine learning and deep learning models;

[0117] The full-chain benchmarking decision module 40 is used for horizontal benchmarking, vertical benchmarking, full-chain benchmarking, weakness diagnosis, optimization recommendation and effect tracking;

[0118] The multi-terminal interaction module 50 is used for output to PCs, mobile devices, large screens, and API interfaces to achieve real-time monitoring, early warning, and decision execution.

[0119] In this embodiment, the full-chain data platform module 20 serves as the system's data hub, performing integrated governance on the multi-source, heterogeneous, and massive data imported by module 10. First, data access and cleaning are performed to remove outliers and invalid data. Then, data from different stages and devices are standardized, unifying measurement units, timestamps, and data formats. Next, through data fusion technology, production data, environmental data, supply chain data, and market data are correlated and integrated, breaking down data silos and forming a full-chain data view based on "one chicken" or "one batch." Finally, blockchain technology is used to store the cleaned and fused key process and quality data, generating irreversible and traceable data fingerprints to ensure the authenticity and credibility of data assets. This module constructs unified, clean, and reliable broiler industry data assets. Data governance improves data quality and usability, data fusion enables cross-stage business insights, and blockchain storage enhances data credibility and product traceability capabilities, providing high-quality, highly consistent input for intelligent models and ensuring the accuracy of decision analysis.

[0120] The multi-terminal interaction module 50 serves as a unified interface layer between the system and users, outputting the functions and data of all the aforementioned modules in a format adapted to different scenarios. For desktop office scenarios, it provides a fully functional PC Web client for in-depth data analysis and report generation. For mobile inspection and real-time management scenarios, it provides a mobile app supporting real-time data viewing, alert reception, and approval processing. For command centers or large-scale display scenarios, it provides a large-screen visualization interface, dynamically displaying the overall macro-level operation of the entire industry chain. Simultaneously, this module opens up core data and service capabilities via API interfaces for integration and access by third-party enterprise systems (such as financial systems and supply chain systems), ensuring seamless integration of data and business flows.

[0121] In a further preferred embodiment of the present invention, the multi-source sensing and acquisition module 10 includes:

[0122] In the breeding process: intelligent chicken scales, temperature and humidity sensors, carbon dioxide sensors, high-definition cameras, RFID tags, and automatic feeding and drinking record equipment;

[0123] Slaughtering and processing: production line visual inspection system, automated cutting line data acquisition points, pre-cooling and quick-freezing temperature monitoring equipment;

[0124] Logistics and warehousing: vehicle-mounted GPS / BeiDou positioning and temperature and humidity recorders, cold storage environment monitoring equipment;

[0125] Management and Market Links: Data interfaces for Enterprise Resource Planning (ERP), Manufacturing Execution System (MES), and Supply Chain Management System (SLM), as well as access interfaces for publicly available market prices, industry standards, and policy and regulatory data sources.

[0126] In this embodiment, the multi-source sensing and acquisition module 10 serves as the system's data access point, deployed across all stages of the broiler industry chain, from breeding to consumption. In the breeding stage, IoT devices such as smart chicken scales, temperature and humidity sensors, high-definition cameras, and RFID are used to collect data on individual chicken weights, group environmental parameters, behavioral videos, and individual identification information in real time and automatically. In the slaughtering and processing stage, a production line vision system automatically acquires carcass quality and grading data, and temperature sensors monitor the temperature control status of key processes such as pre-cooling and quick-freezing. In the logistics and warehousing stage, vehicle-mounted GPS / BeiDou navigation systems, temperature and humidity recorders, and cold storage environment monitoring equipment continuously track logistics trajectories and the cold chain environment. In the management and marketing stage, the system interface connects to internal ERP, MES, and SCM data, and crawls publicly available market prices, industry standards, and policies and regulations, forming an integrated internal and external industry benchmark and market data flow. All data is accessed through standardized protocols, providing raw data fuel for upper-level analysis. This module realizes automated, high-precision, and blind-spot-free collection of multi-dimensional data across the entire broiler industry chain, transforming the traditional production process that relies on manual recording and experience-based judgment into a structured real-time data stream. It fundamentally solves the industry pain points of scattered data sources, inconsistent standards, and poor real-time performance, laying a solid data foundation for subsequent data governance and intelligent analysis.

[0127] In a further preferred embodiment of the present invention, the intelligent model engine module 30 includes:

[0128] The dynamic nutritional requirement model dynamically calculates the optimal nutritional requirements based on parameters such as broiler breed, age, weight, and ambient temperature, and combines them with a raw material price database to generate the lowest cost feed formula.

[0129] The growth and feed conversion ratio prediction model uses historical growth curves, environmental data, and feeding records to predict the date when the flock will reach the target weight, and recommends the most economically efficient slaughter time by taking into account factors such as feed costs, market prices, and pen turnover.

[0130] The disease early warning model is based on video and audio data of chicken behavior, individual body temperature and environmental data to build a multimodal fusion diagnostic model to achieve early warning and intelligent diagnosis of diseases;

[0131] The loss control model predicts carcass quality based on live animal indicators at the end of the breeding period, and uses machine vision to monitor and grade the slaughtering and processing process in real time, optimize the process flow, and reduce losses.

[0132] The whole-chain benchmarking diagnostic model integrates cost data from all stages, conducts input-output analysis, and identifies key points for cost control.

[0133] The risk prediction model integrates historical and real-time market data to predict price trends and assist in sales decisions.

[0134] In this embodiment, the intelligent model engine module 30 is the intelligent core of the system, integrating six professional models. The dynamic nutrition requirement model dynamically calculates the precise daily nutritional needs of broilers based on real-time weight, age, and environmental temperature differences, and optimizes feed formulation in conjunction with a raw material price database. The growth and feed conversion ratio prediction model comprehensively analyzes historical growth curves and environmental data to predict the target weight achievement date, and couples this with a feed cost and market price model to recommend the most economically efficient slaughter time. The disease early warning model integrates multimodal data such as flock behavior videos, sounds, body temperature, and environmental data, and calculates the probability of disease occurrence and provides early warnings several days in advance using logistic regression and LSTM time-series networks. The loss control model quantifies the impact of operational factors in processing, cold chain, and storage on the overall loss rate through regression analysis. The full-chain benchmarking diagnostic model uses Data Envelopment Analysis (DEA) to quantify the relative efficiency of the entire chain operation of the diagnosed entity, identifying weak links with redundant inputs or insufficient outputs. The risk prediction model integrates the time-series characteristics of multiple risks, including disease, cost, price, and losses, to predict the overall short-term risk level. This module transforms industry knowledge and expert experience into a computable and iterative intelligent model, achieving a crucial leap from data to decision-making. It can provide precise nutritional formulas and slaughter recommendations, enable early warning of disease outbreaks, quantify and control losses at each stage, scientifically assess operational efficiency, and predict market and production risks, thereby empowering data-driven refined management and scientific decision-making throughout the entire production process.

[0135] In a further preferred embodiment of the present invention, the dynamic nutritional requirement model includes daily metabolizable energy requirement, daily crude protein requirement, and daily lysine requirement;

[0136] The formula for calculating the daily metabolic energy requirement ME(d) is as follows:

[0137] ;

[0138] Where d is the age of the animals, BW(d) is the actual weight measured on day d, ΔT(d) is the difference between the temperature of the enclosure and the suitable temperature, ADG(d) is the average daily weight gain, ηenv(d) is the comprehensive environmental correction coefficient, and a,b,c,k are the breed-specific fitting coefficients.

[0139] The formula for calculating the daily crude protein requirement CP(d) is as follows:

[0140] ;

[0141] Wherein, CPmaint is the crude protein requirement for maintenance, CPgain is the crude protein requirement per unit weight gain, ε(d) is the stress, immunity, or disease correction term, and UCP is the crude protein utilization rate.

[0142] The formula for calculating the daily lysine requirement (Lys(d)) is as follows:

[0143] ;

[0144] Where α, β, and γ are regression coefficients.

[0145] In a further preferred embodiment of the present invention, the growth and feed conversion ratio prediction model includes a daily cumulative feed conversion ratio function and a multi-factor dynamic FCR prediction function;

[0146] The formula for calculating the cumulative feed conversion ratio for the day is as follows:

[0147] ;

[0148] Wherein, FI(t) is the actual food intake on day t, BW(d) is the predicted weight on day d, and BW(1) is the weight upon entering the enclosure;

[0149] The formula for calculating the multi-factor dynamic FCR prediction function is as follows:

[0150] ;

[0151] in, For the predicted feed conversion ratio on day d, To predict daily weight gain, The temperature inside the building. For the optimal temperature, This refers to the concentration of ammonia gas. For stocking density, For the sake of egg ratio, These are the model fitting coefficients. For random perturbation and health correction terms.

[0152] In a further preferred embodiment of the present invention, the epidemic early warning model includes a core risk probability function, a single-class feature risk quantification function, and an epidemic classification early warning determination function;

[0153] The formula for the core risk probability function is as follows:

[0154] ;

[0155] in, : The overall risk probability of broiler disease on day d, with a value range of [0,1]. The higher the value, the greater the risk.

[0156] Sigmoid activation function This is used to normalize the weighted result to the 0-1 interval and determine the adaptation probability.

[0157] The weight coefficients of the i-th type of risk feature are fitted by training with big data samples and calibrated with industry benchmark data, and the weights sum to 1.

[0158] The standardized risk feature vector of class i on day d covers five core dimensions:

[0159] Environmental characteristics Temperature deviation, ammonia concentration, humidity exceeding the standard, and abnormal ventilation volume;

[0160] Behavioral characteristics : Rate of decrease in feed intake, fluctuation in water consumption, decrease in activity frequency, and rate of recognition of lethargy;

[0161] Physiological and immune characteristics : Antibody titer failure rate, daily weight gain drop rate, and daily mortality rate increase;

[0162] Management characteristics : Overcrowding, failure to meet disinfection frequency standards, and frequency of stress events;

[0163] External public opinion characteristics Regional disease incidence rate, and risk value of disease transmission during transportation / introduction;

[0164] Model bias term, dynamically calibrated based on broiler breed, breeding cycle, and regional disease history;

[0165] The total number of risk features is n=16 in this model, covering all dimensions of disease triggers;

[0166] The formula for the single-class feature risk quantification function is as follows:

[0167] ;

[0168] in, : Measured value of feature on day d;

[0169] : Optimal threshold for features (suitable value for broiler farming);

[0170] Characteristic danger threshold (the critical value that triggers an epidemic);

[0171] The formula for the disease classification and early warning determination function is as follows:

[0172] ;

[0173] in, : The predicted risk probability of the epidemic on day t in the future (t=1,2,3, to achieve short-term advance warning);

[0174] Long Short-Term Memory (LSTM) neural network fitting function to uncover temporal risk patterns;

[0175] The historical risk data time series length is set by taking the risk data from the past 7 days as the time series input; the remaining parameters are consistent with the basic model to ensure that the data are from the same source and the model is linked.

[0176] In a further preferred embodiment of the present invention, the function formula of the loss control model is:

[0177] ;

[0178] in, The overall loss rate (%) of the entire chain in time period t, covering slaughtering and deboning loss, cold chain spoilage loss, and storage damage loss;

[0179] : The model's basic bias term is calibrated based on the industry benchmark loss rate;

[0180] : The weight coefficient of the i-th type of loss influencing factor, with a weight sum of 1;

[0181] Xi(t): The feature vector of standardized loss in time period t, with core dimensions including:

[0182] Processing end: slaughter rate X1(t), deboning process proficiency X2(t), carcass temperature X3(t);

[0183] Cold chain end: cold chain temperature and humidity deviation X4 (t), transportation bump coefficient X5 (t), circulation time deviation X6 (t);

[0184] Warehousing: Warehouse stacking density x7 (t), remaining shelf life percentage of products x8 (t);

[0185] The temporal influence coefficient of historical loss rate (0<β<1) reflects the temporal continuity of loss.

[0186] : Random disturbance term, covering sudden factors.

[0187] In a further preferred embodiment of the present invention, the function formula of the whole-chain benchmarking diagnostic model is as follows: ;

[0188] in, : The overall operational efficiency value of the benchmark entity (such as a certain farm) on day d, with a value of [value missing]. The closer it is to 1, the better the efficiency.

[0189] : The r-th output indicator of the j-th benchmarking unit on day d;

[0190] : The i-th input indicator of the j-th benchmarking unit on day d;

[0191] : Output / input indicator values ​​of the subject being diagnosed;

[0192] The benchmark unit weighting coefficient reflects the reference value of different benchmark units;

[0193] Total number of benchmarking units;

[0194] Number of dimensions for output / input indicators.

[0195] In a further preferred embodiment of the present invention, the function formula of the risk prediction model is:

[0196] ;

[0197] Where k=7: the time series length of historical risk data;

[0198] The time-series characteristics of four types of risks—disease, cost, price, and loss—over the past 7 days;

[0199] Risk level determination mapping function, t=1,2,3 to achieve short-term early warning, the model data is linked with the basic models of disease, loss, nutrition and other basic models.

[0200] In a further preferred embodiment of the present invention, the core functions of the full-chain benchmarking decision module 40 include:

[0201] Dynamic benchmarking analysis: Users define benchmarking targets, and the platform automatically compares the key performance indicators of this session / batch with the benchmarking targets in multiple dimensions, and displays the gaps intuitively in the form of dashboards, trend charts or radar charts;

[0202] Intelligent decision support: Based on benchmarking results and model analysis, the platform automatically generates optimization suggestion reports, providing specific suggested dates for maximizing benefits when hogs are ready for slaughter.

[0203] Full-chain visual traceability: Utilizing blockchain technology, key data from each stage are stored on the chain for evidence, generating an unalterable traceability QR code;

[0204] Early warning and alarm center: Real-time monitoring of key indicators in all aspects. Once the indicators deviate from the preset threshold or the model prediction range, early warning information will be sent to relevant personnel immediately via APP and SMS.

[0205] Knowledge base and expert system: Integrates industry standards, breeding specifications, disease atlases, and medication guidelines, and combines them with a case library to provide online Q&A and decision support for frontline personnel.

[0206] In this embodiment, the full-chain benchmarking decision module 40 provides specific benchmarking analysis and decision support based on the model output of the intelligent model engine module 30 and the post-governance data of the full-chain data platform module 20. The dynamic benchmarking analysis function allows users to flexibly select internal historical data, peer benchmarks, or theoretical optimal values ​​as benchmarking objects. The system automatically calculates the gap in key performance indicators and presents it intuitively through visual charts. The intelligent decision support function integrates the results of growth prediction, cost analysis, and risk warning to automatically generate an analysis report containing specific optimization measures (such as adjusting feeding programs and suggesting slaughter dates). The full-chain visual traceability function calls blockchain-stored data to generate traceability QR codes containing key information from each stage. The early warning and alarm center monitors all model outputs and key indicators in real time, triggering multi-level early warning information pushes once thresholds are exceeded. The knowledge base and expert system provide standard operating procedures and case references for handling specific problems; the full-chain benchmarking decision module 40 transforms the analysis results of the intelligent model into actionable insights and instructions, realizing a closed loop of "monitoring-analysis-decision-execution". It helps managers quickly identify shortcomings through visual benchmarking, drives proactive intervention through automated reporting and early warning, enhances consumer trust through traceability, and improves the capabilities of frontline staff through a knowledge base, ultimately transforming the platform's analytical capabilities into effective operational results in cost reduction, efficiency improvement, and risk control.

[0207] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. 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 benchmarking and decision-making platform for the entire broiler industry chain based on big data and intelligent models, characterized in that, include: Multi-source sensing and acquisition module (10) is used to acquire data on the entire broiler production chain, environment, equipment, supply chain, market and industry benchmarks; The full-chain data platform module (20) is used for data access, cleaning, standardization, integrated governance and blockchain notarization to form a unified data asset; The intelligent model engine module (30) is used to integrate various machine learning and deep learning models; The whole-chain benchmarking decision module (40) is used for horizontal benchmarking, vertical benchmarking, whole-chain benchmarking, weakness diagnosis, optimization recommendation and effect tracking; The multi-terminal interaction module (50) is used for output to PC, mobile, large screen and API interfaces to realize real-time monitoring, early warning and decision execution.

2. The broiler industry chain benchmarking decision-making platform based on big data and intelligent models as described in claim 1, characterized in that, The multi-source sensing and acquisition module (10) includes: In the breeding process: intelligent chicken scales, temperature and humidity sensors, carbon dioxide sensors, high-definition cameras, RFID tags, and automatic feeding and drinking record equipment; Slaughtering and processing: production line visual inspection system, automated cutting line data acquisition points, pre-cooling and quick-freezing temperature monitoring equipment; Logistics and warehousing: vehicle-mounted GPS / BeiDou positioning and temperature and humidity recorders, cold storage environment monitoring equipment; Management and Market Links: Data interfaces for Enterprise Resource Planning (ERP), Manufacturing Execution System (MES), and Supply Chain Management System (SLM), as well as access interfaces for publicly available market prices, industry standards, and policy and regulatory data sources.

3. The broiler industry chain benchmarking decision-making platform based on big data and intelligent models as described in claim 1, characterized in that, The intelligent model engine module (30) includes: The dynamic nutritional requirement model dynamically calculates the optimal nutritional requirements based on parameters such as broiler breed, age, weight, and ambient temperature, and combines them with a raw material price database to generate the lowest cost feed formula. The growth and feed conversion ratio prediction model uses historical growth curves, environmental data, and feeding records to predict the date when the flock will reach the target weight, and recommends the most economically efficient slaughter time by taking into account factors such as feed costs, market prices, and pen turnover. The disease early warning model is based on video and audio data of chicken behavior, individual body temperature and environmental data to build a multimodal fusion diagnostic model to achieve early warning and intelligent diagnosis of diseases; The loss control model predicts carcass quality based on live animal indicators at the end of the breeding period, and uses machine vision to monitor and grade the slaughtering and processing process in real time, optimize the process flow, and reduce losses. The whole-chain benchmarking and diagnostic model integrates cost data from all stages, conducts input-output analysis, and identifies key points for cost control. The risk prediction model integrates historical and real-time market data to predict price trends and assist in sales decisions.

4. The broiler industry chain benchmarking decision-making platform based on big data and intelligent models as described in claim 3, characterized in that, The dynamic nutritional requirement model includes daily metabolizable energy requirement, daily crude protein requirement, and daily lysine requirement. The formula for calculating the daily metabolic energy requirement ME(d) is as follows: ； Where d is the age of the animals, BW(d) is the actual weight measured on day d, ΔT(d) is the difference between the temperature of the enclosure and the suitable temperature, ADG(d) is the average daily weight gain, ηenv(d) is the comprehensive environmental correction coefficient, and a,b,c,k are the breed-specific fitting coefficients. The formula for calculating the daily crude protein requirement CP(d) is as follows: ； Wherein, CPmaint is the crude protein requirement for maintenance, CPgain is the crude protein requirement per unit weight gain, ε(d) is the stress, immunity, or disease correction term, and UCP is the crude protein utilization rate. The formula for calculating the daily lysine requirement (Lys(d)) is as follows: ； Where α, β, and γ are regression coefficients.

5. The broiler industry chain benchmarking decision-making platform based on big data and intelligent models according to claim 3, characterized in that, The growth and feed conversion ratio prediction model includes a daily cumulative feed conversion ratio function and a multi-factor dynamic FCR prediction function. The formula for calculating the cumulative feed conversion ratio function for the day is as follows: ； Wherein, FI(t) is the actual food intake on day t, BW(d) is the predicted weight on day d, and BW(1) is the weight upon entering the enclosure; The calculation formula for the multi-factor dynamic FCR prediction function is as follows: ； in, For the predicted feed conversion ratio on day d, To predict daily weight gain, The temperature inside the building. For the optimal temperature, This refers to the concentration of ammonia gas. For stocking density, For the sake of egg ratio, These are the model fitting coefficients. For random perturbation and health correction terms.

6. The broiler industry chain benchmarking decision-making platform based on big data and intelligent models according to claim 3, characterized in that, The disease early warning model includes a core risk probability function, a single-class feature risk quantification function, and a disease classification early warning judgment function. The formula for the core risk probability function is as follows: ； in, : The overall risk probability of broiler disease on day d, with a value range of [0,1]. The higher the value, the greater the risk. Sigmoid activation function This is used to normalize the weighted result to the 0-1 interval and determine the adaptation probability. The weight coefficients of the i-th type of risk feature are fitted by training with big data samples and calibrated with industry benchmark data, and the weights sum to 1. The standardized risk feature vector of class i on day d covers five core dimensions: Environmental characteristics Temperature deviation, ammonia concentration, humidity exceeding the standard, and abnormal ventilation volume; Behavioral characteristics : Rate of decrease in feed intake, fluctuation in water consumption, decrease in activity frequency, and rate of recognition of lethargy; Physiological and immune characteristics : Antibody titer failure rate, daily weight gain drop rate, and daily mortality rate increase; Management characteristics : Overcrowding, failure to meet disinfection frequency standards, and frequency of stress events; External public opinion characteristics Regional disease incidence rate, and risk value of disease transmission during transportation / introduction; Model bias term, dynamically calibrated based on broiler breed, breeding cycle, and regional disease history; The total number of risk features is n=16 in this model, covering all dimensions of disease triggers; The formula for the single-class feature risk quantification function is as follows: ； in, : Measured value of feature on day d; : Optimal threshold for features (suitable value for broiler farming); Characteristic danger threshold (the critical value that triggers an epidemic); The formula for the disease classification and early warning determination function is as follows: ； in, : The predicted risk probability of the epidemic on day t in the future (t=1,2,3, to achieve short-term advance warning); Long Short-Term Memory (LSTM) neural network fitting function to uncover temporal risk patterns; The historical risk data time series length is set by taking the risk data from the past 7 days as the time series input; the remaining parameters are consistent with the basic model to ensure that the data are from the same source and the model is linked.

7. The broiler industry chain benchmarking decision-making platform based on big data and intelligent models according to claim 3, characterized in that, The function formula of the loss control model is: ； in, The overall loss rate (%) of the entire chain in time period t, covering slaughtering and deboning loss, cold chain spoilage loss, and storage damage loss; : The model's basic bias term is calibrated based on the industry benchmark loss rate; : The weight coefficient of the i-th type of loss influencing factor, with a weight sum of 1; Xi(t): The feature vector of standardized loss in time period t, with core dimensions including: Processing end: slaughter rate X1(t), deboning process proficiency X2(t), carcass temperature X3(t); Cold chain end: cold chain temperature and humidity deviation X4 (t), transportation bump coefficient X5 (t), circulation time deviation X6 (t); Warehousing: Warehouse stacking density x7 (t), remaining shelf life percentage of products x8 (t); The temporal influence coefficient of historical loss rate (0<β<1) reflects the temporal continuity of loss. : Random disturbance term, covering sudden factors.

8. The broiler industry chain benchmarking decision-making platform based on big data and intelligent models according to claim 3, characterized in that, The function formula for the whole-chain benchmarking diagnostic model is: ； in, : The overall operational efficiency value of the benchmark entity (such as a certain farm) on day d, with a value of [value missing]. The closer it is to 1, the better the efficiency. : The r-th output indicator of the j-th benchmarking unit on day d; : The i-th input indicator of the j-th benchmarking unit on day d; : Output / input indicator values ​​of the subject being diagnosed; The benchmark unit weighting coefficient reflects the reference value of different benchmark units; Total number of benchmarking units; Number of dimensions for output / input indicators.

9. A broiler industry chain benchmarking decision-making platform based on big data and intelligent models as described in claim 3, characterized in that, The function formula for the risk prediction model is: ； Where k=7: the time series length of historical risk data; The time-series characteristics of four types of risks—disease, cost, price, and loss—over the past 7 days; Risk level determination mapping function, t=1,2,3 to achieve short-term advanced early warning, the model data is from the same source and linked with the basic models of epidemic, loss, nutrition and other factors.

10. A broiler industry chain benchmarking decision-making platform based on big data and intelligent models as described in claim 1, characterized in that, The core functions of the full-chain benchmarking decision module (40) include: Dynamic benchmarking analysis: Users define benchmarking targets, and the platform automatically compares the key performance indicators of this session / batch with the benchmarking targets in multiple dimensions, and displays the gaps intuitively in the form of dashboards, trend charts or radar charts; Intelligent decision support: Based on benchmarking results and model analysis, the platform automatically generates optimization suggestion reports, providing specific suggested dates for maximizing benefits when hogs are ready for slaughter. Full-chain visual traceability: Utilizing blockchain technology, key data from each stage are stored on the chain for evidence, generating an unalterable traceability QR code; Early warning and alarm center: Real-time monitoring of key indicators in all aspects. Once the indicators deviate from the preset threshold or the model prediction range, early warning information will be sent to relevant personnel immediately via APP and SMS. Knowledge base and expert system: Integrates industry standards, breeding specifications, disease atlases, and medication guidelines, and combines them with a case library to provide online Q&A and decision support for frontline personnel.

Images