Feed additive ratio control method and system based on environmental perception

Abstract

The invention discloses a feed additive ratio control method and system based on environmental perception. The method comprises the steps that breeding environment, feed characteristics, equipment operation and animal physiological parameters are collected, and a high-quality parameter set is generated after preprocessing; calculating an environment stress index based on a multi-factor data fusion algorithm and analyzing a change trend; combining historical and current data, using time sequence prediction and a machine learning algorithm to predict additive requirements and generating an optimal proportioning strategy; a control instruction is generated according to the strategy, and the additive conveying and spraying process is regulated and controlled; and a matching strategy and control parameters are corrected in real time through closed-loop feedback, so that accurate regulation and control are realized. The device can dynamically adapt to environmental changes, reduce animal stress, and improve mixing uniformity and nutrition retention effect.

Description

Technical Field

[0001] This invention relates to the intersection of feed processing and aquaculture environment control, specifically to the field of precise ratio control technology for feed additives based on multi-source sensing and intelligent algorithms. Background Technology

[0002] In modern animal husbandry and feed processing, the control of feed additive ratios is a key link in ensuring balanced feed nutrition and improving animal husbandry efficiency. Its core is to precisely adjust the mixing ratio of additives and basic feed according to the actual needs of the farming scenario, adapting it to the animal's physiological state, environmental conditions, and feed characteristics, ultimately achieving the goals of reducing animal stress and improving feed utilization. With the development of large-scale and intelligent farming, traditional additive ratio control methods relying on manual experience or single-parameter adjustments are no longer sufficient to meet the demands for high-precision, dynamic control. The industry's need for ratio control technologies that integrate multi-source sensing and intelligent algorithms is becoming increasingly urgent.

[0003] Currently, existing feed additive formulation technologies suffer from three main shortcomings, which significantly differ from the core invention of this solution. First, parameter collection is limited in scope and data reliability is low. Existing technologies typically collect only a few parameters, such as temperature and humidity in the livestock shed or basic feed components, failing to incorporate animal physiological parameters (e.g., heart rate, feed intake). Furthermore, data processing relies solely on simple filtering, which cannot effectively eliminate sensor noise and drift interference, resulting in insufficient accuracy of input data for subsequent formulation calculations and an inability to accurately reflect the overall state of the livestock farming environment. Second, environmental stress assessment is vague and formulation lacks predictive power. Existing technologies rely heavily on manual observation or single-parameter threshold judgments for assessing "environmental stress," lacking quantitative stress assessment indicators and failing to accurately measure the degree of environmental stress on animals. Simultaneously, formulation adjustments are often "passive responses," only made after animal stress or abnormal feed quality is detected, without employing time-series predictive models to anticipate additive needs. This leads to formulation adjustments lagging behind dynamic changes in the environment and animal condition, resulting in poor stress mitigation. Third, the control execution and feedback mechanisms are inadequate. In existing technologies, the spraying pressure of additives and the material conveying speed are mostly controlled independently, without establishing a dynamic adaptation relationship between the two. This often leads to poor mixing uniformity due to mismatch between pressure and speed. Furthermore, there is a lack of closed-loop feedback throughout the entire process. Only a few solutions monitor mixing uniformity and do not combine multi-dimensional effect parameters such as nutrient retention rate and changes in animal feed intake to dynamically correct the ratio strategy. This makes it impossible to continuously optimize the ratio accuracy and adapt to the differentiated needs of different breeding scenarios.

[0004] Therefore, we need to develop an additive ratio control technology that integrates precise acquisition and intelligent processing of multi-source parameters, quantitative environmental stress assessment, and demand prediction functions to overcome the above-mentioned technical problems. Summary of the Invention

[0005] In view of the defects and problems of existing technologies, this invention aims to provide a feed additive ratio control method based on environmental perception. Through precise processing of multi-source parameters, quantitative assessment of stress, demand prediction and closed-loop feedback, the method can achieve precise ratio control, reduce animal stress and improve feed utilization.

[0006] The solution to the technical problem of this invention is as follows: a feed additive ratio control method based on environmental perception is adopted, the method comprising the following steps: S1, real-time acquisition of breeding environment parameters, feed characteristic parameters, material conveying equipment operation parameters, and animal physiological parameters through the multi-parameter acquisition module, and preprocessing the acquired parameters to remove data noise and drift interference to generate a high-quality parameter set; S2, based on the high-quality parameter set, a quantitative value comprehensively representing the degree of environmental stress, namely the environmental stress index, is calculated by the environmental stress intelligent assessment module using a multi-factor data fusion algorithm, and the environmental change trend is analyzed based on historical and current environmental stress indices; S3, based on historical and current environmental stress indices and system operating status parameters, the future time is predicted by the time series prediction model in the intelligent prediction decision module. The additive demand baseline value for a segment is determined, and an optimal ratio strategy is generated through machine learning algorithms in conjunction with real-time collected multi-source parameters; S4. Based on the predicted additive demand baseline value and optimal ratio strategy, control instructions for real-time adjustment of the adaptive control execution module are generated through the optimization control algorithm in the intelligent prediction decision module; S5. The adaptive control execution module adjusts the amount of additive delivered from the additive storage unit to the buffer component according to the control instructions, and synchronously drives the pressure supply unit to generate adaptive pressure as the material conveying equipment operates, so that the additive is accurately sprayed onto the feed at the discharge port of the material conveying equipment through the spraying unit; S6. The mixing effect parameters of the additive and feed are collected in real time through the closed-loop feedback optimization module, and the optimal ratio strategy and control parameters are corrected based on the mixing effect parameters to form a closed-loop monitoring system.

[0007] Preferably, in step S1, the preprocessing includes the following specific steps: S11, using an adaptive Kalman filter algorithm to filter out noise from the acquired raw parameters, and dynamically adjusting the filter gain according to the parameter type; S12, correcting parameter drift using a linear trend compensation algorithm, fitting the parameter change trend within 5-10 consecutive sampling periods, and eliminating systematic drift caused by sensor aging or environmental interference; S13, standardizing the parameters after noise filtering and drift correction, mapping parameters of different dimensions to the [0,1] interval, and generating the high-quality parameter set.

[0008] Preferably, in step S2, the multi-factor data fusion algorithm includes a weighted fusion sub-algorithm and an attention mechanism sub-algorithm, comprising the following steps: S21, assigning basic weights to each parameter in the high-quality parameter set, wherein the weight of harmful gas concentration in the breeding environment parameters accounts for 30%-40%, the weight of heart rate / feed intake in animal physiological parameters accounts for 25%-35%, and the weight of feed characteristic parameters and equipment operation parameters accounts for 25%-45%; S22, dynamically adjusting the basic weights through the attention mechanism sub-algorithm, increasing the weight of parameters that currently exceed a preset threshold by 10%-20%; S23, performing weighted calculations on each parameter based on the adjusted weights to obtain the environmental stress index.

[0009] Preferably, in step S3, the time-series prediction model is a long short-term memory network or a Transformer model, and the prediction includes the following steps: S31, using historical environmental stress index, additive usage, and animal growth status data from the past 3-7 days as training samples to train the time-series prediction model; S32, inputting the current environmental stress index and predicted environmental parameters for the next 2-4 hours, and outputting the additive requirement baseline value for the next 5-10 minutes through the trained time-series prediction model.

[0010] Preferably, in step S3, the machine learning algorithm is a fuzzy neural network or a deep reinforcement learning algorithm: if a fuzzy neural network algorithm is used, a high-quality parameter set and an environmental stress index are used as input features, and "minimizing animal stress + optimizing additive cost" is used as the output objective to construct an input-output fuzzy mapping relationship and generate an optimal formulation strategy; if a deep reinforcement learning algorithm is used, "the reduction rate of environmental stress index" and "the improvement rate of feed utilization" are used as reward functions, and formulation decision strategies adapted to different scenarios are obtained through multiple rounds of iterative training.

[0011] Preferably, in step S5, the synchronous drive pressure supply unit generates an adaptive pressure, and establishes a positive correlation between the conveyor shaft speed of the material conveying equipment and the output pressure of the pressure supply unit through a parameter correlation model; the conveyor shaft speed is collected in real time, the target pressure value is calculated through the correlation model, and the extrusion frequency of the elastic pressure component of the pressure supply unit is driven to match the output pressure value, thereby ensuring that the amount of additive sprayed is matched with the amount of feed conveyed.

[0012] Preferably, in step S6, the mixing effect parameters include additive mixing uniformity, feed nutrient retention rate, and animal short-term feed intake change rate; the specific logic of the feedback correction is as follows: if the mixing uniformity RSD value is >5%, then reduce the additive spraying flow rate by 10%-15% and increase the output pressure of the pressure supply unit by 5%-8%; if the vitamin activity retention rate is <85%, then adjust the atomized particle diameter of the spraying unit, and at the same time correct the optimal ratio strategy in step S3, increasing the proportion of vitamin additives by 2%-5%.

[0013] An intelligent feed additive precision proportioning control system, the system being used to execute the feed additive proportioning control method based on environmental perception as described in any one of claims 1-7, comprising: a multi-parameter acquisition module for real-time acquisition of breeding environment parameters, feed characteristic parameters, material conveying equipment operating parameters, and animal physiological parameters, the multi-parameter acquisition module including a distributed sensor group; a data processing and fusion module, signal-connected to the multi-parameter acquisition module, for performing preprocessing, feature extraction, and quality assessment on the acquired multi-source parameters; an intelligent environmental stress assessment module, signal-connected to the data processing and fusion module, for calling a multi-factor data fusion algorithm, calculating an environmental stress index based on a high-quality parameter set, and outputting environmental change trend analysis results; and an intelligent prediction and decision-making module, and... The environmental stress intelligent assessment module is signal-connected and incorporates a time-series prediction model and machine learning algorithm to predict additive demand baseline values ​​and generate optimal mixing strategies. It also generates control commands through an optimized control algorithm. The adaptive control execution module, signal-connected to the intelligent prediction and decision module, includes an additive metering submodule, a pressure control submodule, and a spraying control submodule. It precisely regulates the additive delivery rate, pressure supply unit output pressure, and spraying parameters according to control commands. The closed-loop feedback optimization module, signal-connected to both the adaptive control execution module and the intelligent prediction and decision module, includes a mixing effect detection sensor and a correction algorithm unit. It collects mixing effect parameters in real time and feeds back correction commands to the intelligent prediction and decision module to dynamically adjust the mixing strategy and control parameters.

[0014] The beneficial effects of this invention are: 1. By sensing animal physiological parameters and quantitatively assessing stress index, combined with demand prediction and dynamic ratio adjustment, it can accurately adapt to changes in environmental stress, effectively reduce animal stress response, and ensure the stability of animal physiological state.

[0015] 2. Multi-factor data fusion algorithms avoid evaluation bias caused by parameter weight imbalance, while time-series prediction and machine learning algorithms solve the problem of multi-parameter coupling, enabling additive formulations to accurately match feed characteristics, equipment status and animal needs, and significantly improving mixing uniformity and nutrient retention.

[0016] 3. The closed-loop control throughout the entire process reduces human intervention, the adaptive execution mechanism ensures that pressure and delivery speed are matched, and the feedback optimization module can dynamically correct deviations, enabling the system to operate stably in different breeding scenarios and equipment wear conditions, significantly improving adaptability and reliability.

[0017] 4. The optimal formulation strategy aims to optimize costs and avoid excessive waste of additives. At the same time, it improves nutrient retention, fully utilizes the nutritional value of feed, significantly increases feed utilization, and balances breeding efficiency with resource conservation. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the operational logic of the control method of the present invention; Figure 2 This is a system module interaction logic diagram; Figure 3 This is a schematic diagram of the distributed deployment of multi-parameter acquisition modules; Figure 4 This is a comparison chart of the parameter preprocessing process and its effects; Figure 5 This is a flowchart illustrating the logic of calculating the environmental stress index. Figure 6 It is a flowchart for demand forecasting and allocation decisions; Figure 7 This is the control block diagram of the adaptive execution module; Figure 8 This is a block diagram of closed-loop feedback optimization logic; Figure 9 This invention is based on a diagram showing the module relationships and core parameter distribution of the system. Detailed Implementation

[0019] The following detailed description of the environmental perception-based feed additive ratio control method and the corresponding intelligent feed additive precision ratio control system of the present invention, combined with specific implementation process, aims to clearly present the technical implementation path of the present invention, highlight the innovation and logic of the technical means, rather than simply emphasizing the technical effect.

[0020] In this embodiment, the intelligent feed additive precision ratio control system (such as...) Figure 2 and Figure 9 (As shown) First, complete the signal connection and initialization of each module to ensure stable signal transmission between the multi-parameter acquisition module, data processing and fusion module, environmental stress intelligent assessment module, intelligent prediction and decision-making module, adaptive control execution module, and closed-loop feedback optimization module. After the built-in algorithms and preset parameters of each module are loaded, the feed additive ratio control method based on environmental perception is then started. The system hardware-parameter-algorithm relationship is shown in the table below.

[0021]

[0022] Figure 9 In this process, raw parameters of the breeding shed / animals / feed / equipment (such as ammonia concentration and feed moisture) are collected by distributed sensors and transmitted to the data processing and fusion module. Noise is dynamically filtered out by adaptive Kalman filtering (such as high-gain filtering for ammonia concentration and low-gain filtering for feed parameters), and sensor drift is corrected by linear trend compensation (based on fitting over 5-10 sampling periods). Finally, the parameters are standardized into a high-quality parameter set in the [0,1] interval. The high-quality parameter set is transmitted to the environmental stress intelligent assessment module, where a weighted fusion algorithm is used to assign basic weights (30%-40% for harmful gases, 25%-35% for animal physiology, and 25%-45% for feed / equipment). Then, an attention mechanism is used to increase the weight of parameters exceeding the limit (such as adding 10%-20% when ammonia > 20ppm), and an environmental stress index of 0-10 is calculated. The index is combined with historical data, and the LSTM / Transformer model of the intelligent prediction and decision-making module predicts the additive demand for the next 5-10 minutes. Finally, a "stress minimization + cost optimization" formulation strategy is generated through fuzzy neural network / DRL. The adaptive control execution module translates the mixing strategy into instructions: adjusting the metering pump speed (controlling the amount of additive delivered), adjusting the pressure supply unit (for every 10% increase in conveyor shaft speed, the pressure increases by 8%-12% simultaneously), and optimizing the nozzle opening; at the same time, the closed-loop feedback optimization module collects mixing effect parameters through downstream sensors (e.g., when RSD > 5%, the instruction is to reduce the spray flow rate by 10%-15% and increase the pressure by 5%-8%), and the MPC algorithm corrects the mixing / control parameters to form a closed loop.

[0023] Feed additive ratio control methods such as Figure 1 As shown, the process involves: S1 - Real-time acquisition and preprocessing of multi-source data; S2 - Intelligent calculation and assessment of environmental stress index; S3 - Additive demand prediction based on time-series forecasting; S4 - Optimal proportion decision based on multi-parameter fusion; S5 - Adaptive control command generation and execution; S6 - Real-time monitoring and feedback of mixing effect; S7 - Dynamic optimization and calibration of control parameters; and S8 - Continuous learning and improvement of system performance. The specific steps are as follows.

[0024] First, the multi-parameter acquisition module is activated, and its distributed sensor array begins operation. Sensors deployed in key areas of the livestock shed collect real-time environmental parameters, including temperature and humidity, concentrations of harmful gases such as ammonia and hydrogen sulfide, and light intensity. Sensors deployed along the feed conveying path collect feed characteristic parameters, including moisture content, crude protein content, and particle size. Sensors deployed on material conveying equipment collect equipment operating parameters, primarily the rotational speed of the conveyor shaft and the flow rate at the discharge port. Sensors deployed in the animal activity area collect animal physiological parameters, such as heart rate, feed intake, and activity frequency. During the acquisition process, each sensor continuously acquires data according to a preset sampling period, such as... Figure 3 As shown, this ensures the real-time and continuous nature of parameter acquisition.

[0025] After acquiring the raw parameters, they are transmitted to the data processing and fusion module for preprocessing. The first step is noise filtering using an adaptive Kalman filter algorithm. This algorithm dynamically adjusts the filter gain based on the parameter type. For environmental parameters such as the concentration of harmful gases that are easily affected by external interference, the filter gain is appropriately increased to enhance noise suppression capabilities. For relatively stable feed characteristic parameters, the filter gain is decreased to reduce the loss of useful information. Through this dynamic adjustment, random noise in the raw parameters is effectively filtered out.

[0026] After noise filtering, the parameter drift correction stage begins, employing a linear trend compensation algorithm. This algorithm fits the parameter change trend over 5-10 consecutive sampling periods, identifying systematic drift trends caused by sensor aging or environmental interference. Subsequently, it performs reverse compensation on the parameters based on the fitted trend line, thereby eliminating parameter deviations caused by drift and ensuring parameter accuracy.

[0027] Finally, standardization is performed. Since different types of parameters have different dimensions (e.g., temperature and humidity have completely different dimensions than harmful gas concentration), directly using them for subsequent calculations would affect the accuracy of the results. Therefore, all parameters after noise filtering and drift correction are mapped to the [0,1] interval, providing a unified comparison benchmark for different parameters and generating a high-quality parameter set, thus providing a reliable data foundation for subsequent environmental stress assessment. The calculation of the environmental stress index comprehensively considers the synergistic effects of multiple environmental factors, including temperature, humidity, gas concentration, noise, and light intensity, and uses a weighted fusion algorithm to generate quantitative assessment results.

[0028] Simultaneously, the data processing and fusion module also performs feature extraction and quality assessment on high-quality parameter sets, extracting derived features such as parameter change rate and threshold exceedance frequency to assist subsequent analysis; it also removes outliers from the dataset to ensure a parameter pass rate ≥ 98%, such as... Figure 4 As shown, this further ensures data quality.

[0029] Secondly, the environmental stress index is calculated and its trend is analyzed. The data processing and fusion module transmits the generated high-quality parameter set to the intelligent environmental stress assessment module. This module calls a multi-factor data fusion algorithm to calculate the environmental stress index and analyze environmental change trends. The multi-factor data fusion algorithm includes a weighted fusion sub-algorithm and an attention mechanism sub-algorithm. First, the weighted fusion sub-algorithm assigns basic weights to each parameter in the high-quality parameter set.

[0030] Considering the varying degrees of impact of different parameters on animal environmental stress, the concentration of harmful gases (ammonia, hydrogen sulfide) in the breeding environment directly affects animal respiratory health and contributes significantly to environmental stress; therefore, its basic weighting is set at 30%-40%. Animal physiological parameters such as heart rate and feed intake directly reflect the animal's current physiological state and are closely related to stress levels; their basic weighting is set at 25%-35%. While feed characteristic parameters and equipment operating parameters also affect environmental stress, their impact is relatively indirect; their basic weighting is set at 25%-45%. This weighting allocation initially reflects the differences in importance among the various parameters.

[0031] Next, the attention mechanism sub-algorithm is activated. This algorithm monitors in real time whether the values ​​of each parameter exceed the preset threshold. When a parameter exceeds the threshold, such as ammonia concentration > 20 ppm, it indicates that the parameter has had a significant stress effect on the animal. At this time, the weight of the parameter is automatically increased by 10%-20%, further strengthening the role of key stress factors in the index calculation and making the weight allocation more in line with the actual stress situation.

[0032] After weight adjustment, the parameters are weighted and calculated to obtain an environmental stress index ranging from 0 to 10. A higher index indicates a stronger degree of environmental stress. Simultaneously, the intelligent environmental stress assessment module retrieves past environmental stress indices from historical data storage units and compares them with the currently calculated index. A trend fitting algorithm identifies the changing trend of environmental stress—whether it is increasing, decreasing, or stable. Figure 5 As shown, this provides a basis for predicting the demand trends of additives in the future.

[0033] Then, the additive demand forecasting and optimal ratio strategy generation are performed. The environmental stress intelligent assessment module transmits the calculated historical and current environmental stress indices, as well as system operating status parameters (such as past additive usage, equipment operating stability parameters, etc.) to the intelligent prediction and decision-making module. This module completes the additive demand forecasting and optimal ratio strategy generation through the built-in time series prediction model and machine learning algorithm.

[0034] In this process, the time-series prediction model is first activated. In this embodiment, a Long Short-Term Memory (LSTM) network or a Transformer model can be selected. Before using the model, it is trained by inputting historical environmental stress index, additive usage, and animal growth status data from the past 3-7 days as training samples. The model parameters are adjusted iteratively to enable the model to have accurate predictive capabilities. The time-series prediction model uses a Long Short-Term Memory (LSTM) network or Gated Recurrent Unit (GRU) deep learning algorithm, which can capture long-term dependencies in environmental changes.

[0035] After training, the current environmental stress index and the predicted environmental parameters for the next 2-4 hours derived from real-time parameter trends are input into the model. The model learns from historical data patterns and analyzes current trends to output the additive requirement baseline value for the next 5-10 minutes. This baseline value includes the total amount of additives added and the proportion range of core additives, providing a basic reference for subsequent formulation of proportioning strategies.

[0036] Subsequently, the intelligent prediction and decision-making module invokes machine learning algorithms to generate the optimal allocation strategy by combining real-time collected multi-source parameters (i.e., a high-quality parameter set). The machine learning algorithm can be either a fuzzy neural network (FNN) or a deep reinforcement learning (DRL) algorithm.

[0037] If a fuzzy neural network algorithm is chosen, a high-quality parameter set and environmental stress index can be used as input features. A fuzzy rule base can be constructed with "minimizing animal stress + optimizing additive cost" as the output objective. Through the nonlinear mapping capability of the neural network, a fuzzy mapping relationship between the input features and the output objective can be established, thereby generating an optimal ratio strategy adapted to the current environment and animal state, and clarifying the specific addition ratios of various additives. Figure 6 As shown. The machine learning algorithm includes a reinforcement learning algorithm, which can continuously optimize the control strategy through trial and error, thereby improving the system's adaptability and control accuracy.

[0038] If a deep reinforcement learning algorithm is selected, and the "reduction rate of environmental stress index" and "improvement rate of feed utilization" are set as reward functions, the algorithm will learn the reward value change pattern under different combinations of environmental parameters and additive ratios during continuous iterative training, gradually optimize the ratio decision, and finally obtain the optimal ratio strategy adapted to different breeding scenarios, ensuring that feed utilization efficiency is improved while reducing animal stress.

[0039] During the control command generation and adaptive execution process, the intelligent predictive decision-making module, based on the predicted additive demand baseline and optimal proportioning strategy, calls the built-in optimization control algorithm to further optimize the proportioning strategy, transforming it into control commands that can directly drive the adaptive control execution module. These commands cover specific control parameters such as additive delivery rate, pressure supply unit output pressure, and spraying unit operating parameters. The closed-loop feedback optimization module employs a model predictive control (MPC) algorithm, enabling optimal control under system constraints.

[0040] After the control command is generated, it is transmitted to the adaptive control execution module, which executes the corresponding operation according to the command. First, it regulates the metering pump in the additive metering submodule. Based on the delivery requirements of various additives in the command, it precisely controls the operating speed of the metering pump to deliver the additives from the additive storage unit to the buffer component, ensuring that the delivery amount of each additive conforms to the optimal ratio strategy.

[0041] Meanwhile, the adaptive control execution module drives the pressure supply unit to generate adaptive pressure as the material conveying equipment operates. This is achieved through a parameter correlation model, which pre-establishes a positive correlation between the conveyor shaft speed of the material conveying equipment and the output pressure of the pressure supply unit. For example, for every 10% increase in speed, the pressure increases by 8%-12%.

[0042] The module collects the conveyor shaft speed in real time, inputs the speed data into the parameter association model, calculates the target pressure value under the current working conditions, and then drives the compression frequency of the elastic pressure component (hemispherical air bladder) in the pressure supply unit. By adjusting the compression frequency, the pressure intensity of the air bladder on the additive is changed, so that the output pressure of the pressure supply unit accurately matches the target pressure value, ensuring that the amount of additive sprayed is compatible with the amount of feed conveyed, and avoiding excessive or insufficient additive spraying due to mismatch between pressure and speed.

[0043] Finally, the adaptive control execution module controls the spraying control submodule, adjusting parameters such as nozzle opening to ensure that the additive is precisely sprayed into the feed at the discharge port of the material conveying equipment via the spraying unit in a suitable manner. Figure 7 As shown, this achieves the initial mixing of the additive with the feed.

[0044] During the mixing effect monitoring and closed-loop feedback optimization, the closed-loop feedback optimization module works continuously during the mixing process of additives and feed. The mixing effect detection sensor included in the module collects mixing effect parameters in real time, mainly including additive mixing uniformity (expressed as RSD value), feed nutrient retention rate (such as vitamin activity retention rate), and animal short-term feed intake change rate.

[0045] After collecting the mixing effect parameters, the closed-loop feedback optimization module analyzes the parameters, determines whether the current mixing effect meets the preset requirements, and generates correction instructions based on the analysis results. These instructions are then fed back to the intelligent prediction and decision-making module to dynamically adjust the optimal mixing strategy and control parameters, thus forming a closed-loop monitoring system.

[0046] If the RSD value of the mixing uniformity is detected to be greater than 5%, it indicates that the current additive and feed are not mixed evenly enough, which may result in some feeds having excessively high additive content and others having excessively low content, affecting the animal's intake. At this time, the generated correction instruction will require reducing the additive spraying flow rate by 10%-15%, reducing the total amount of additive sprayed per unit time, while increasing the output pressure of the pressure supply unit by 5%-8%, enhancing the atomization and diffusion ability of the additive, thereby improving the mixing uniformity.

[0047] If the vitamin activity retention rate is detected to be <85%, it indicates that the vitamin additives suffered significant activity loss due to environmental factors (such as pressure and temperature) during the spraying and mixing process, which will affect the nutritional value of the feed. The corresponding correction instruction will require adjusting the atomized particle diameter of the spraying unit, increasing it by 10%-20% to reduce the contact area between vitamins and the external environment during atomization, thus reducing activity loss; simultaneously, it will correct the optimal formulation strategy generated in step S3 of the intelligent prediction and decision-making module, increasing the proportion of vitamin additives by 2%-5% to compensate for the nutritional gap caused by activity loss. Figure 8 As shown. Through such closed-loop feedback optimization, the formulation strategy and control parameters are continuously adjusted to ensure that the feed additive ratio is always in the optimal state, thus guaranteeing feed quality and animal husbandry results.

[0048] The above embodiments detail the technical implementation process of the present invention. Through a series of technical means such as accurate acquisition and processing of multi-source parameters, quantitative assessment of environmental stress, prediction and intelligent formulation of additive requirements, adaptive execution and closed-loop optimization, the present invention solves the problems of low parameter reliability, fuzzy stress assessment, lack of predictability in formulation, lack of adaptability in control, and lack of closed-loop feedback in traditional feed additive formulation methods. It achieves the precision, dynamism, and intelligence of feed additive formulation, and the various technical means are logically coherent and closely related, together forming a complete technical solution.

[0049] It should be noted that the above embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In practical applications, the parameters and weight ratios of each algorithm can be appropriately adjusted according to specific breeding scenarios, animal species, and other factors. As long as the core technical concepts and technical means of the present invention are adopted, they all fall within the protection scope of the present invention.

Claims

1. A method for controlling the proportion of feed additives based on environmental perception, characterized in that, The method includes the following steps: S1. The multi-parameter acquisition module collects breeding environment parameters, feed characteristic parameters, material conveying equipment operation parameters and animal physiological parameters in real time, and preprocesses the collected parameters to remove data noise and drift interference, generating a high-quality parameter set. S2. Based on the high-quality parameter set, the environmental stress intelligent assessment module uses a multi-factor data fusion algorithm to calculate a quantitative value that comprehensively represents the degree of environmental stress, namely the environmental stress index, and analyzes the trend of environmental change based on historical and current environmental stress indices. S3. Based on historical and current environmental stress indices and system operating status parameters, the time-series prediction model in the intelligent prediction and decision-making module is used to predict the baseline value of additive demand for future periods, and combined with real-time collected multi-source parameters, the optimal ratio strategy is generated through machine learning algorithms. S4. Based on the predicted additive demand baseline value and optimal ratio strategy, the intelligent prediction decision module generates control instructions for real-time adjustment of the adaptive control execution module through the optimization control algorithm in the intelligent prediction decision module. S5. The adaptive control execution module adjusts the amount of additive delivered from the additive storage unit to the buffer component according to the control command, and synchronously drives the pressure supply unit to generate adaptive pressure as the material conveying equipment operates, so that the additive is accurately sprayed onto the feed at the discharge port of the material conveying equipment through the spraying unit. S6. The mixing effect parameters of the additive and feed are collected in real time through the closed-loop feedback optimization module, and the optimal ratio strategy and control parameters are corrected based on the mixing effect parameters to form a closed-loop monitoring.

2. The feed additive ratio control method based on environmental perception according to claim 1, characterized in that, In step S1, the preprocessing includes the following specific steps: S11, using an adaptive Kalman filter algorithm to filter out noise from the acquired raw parameters, and dynamically adjusting the filter gain according to the parameter type; S12, correcting parameter drift using a linear trend compensation algorithm, fitting the parameter change trend within 5-10 consecutive sampling periods, and eliminating systematic drift caused by sensor aging or environmental interference; S13, standardizing the parameters after noise filtering and drift correction, mapping parameters of different dimensions to the [0,1] interval, and generating the high-quality parameter set.

3. The feed additive ratio control method based on environmental perception according to claim 1 or 2, characterized in that, In step S2, the multi-factor data fusion algorithm includes a weighted fusion sub-algorithm and an attention mechanism sub-algorithm, comprising the following steps: S21. Assign basic weights to each parameter in the high-quality parameter set, with the weight of harmful gas concentration in the breeding environment parameters accounting for 30%-40%, the weight of heart rate / feed intake in animal physiological parameters accounting for 25%-35%, and the weight of feed characteristic parameters and equipment operation parameters accounting for 25%-45%. S22. Dynamically adjust the basic weights through the attention mechanism sub-algorithm, and increase the weight ratio of parameters that exceed the preset threshold by 10%-20%; S23. The environmental stress index is obtained by weighting each parameter based on the adjusted weights.

4. The feed additive ratio control method based on environmental perception according to any one of claims 1-3, characterized in that, In step S3, the time-series prediction model is a long short-term memory network or a Transformer model, and the prediction includes the following steps: S31. Use historical data on environmental stress index, additive usage, and animal growth status over the past 3-7 days as training samples to train the time-series prediction model. S32. Input the current environmental stress index and the predicted values ​​of environmental parameters for the next 2-4 hours. The trained time-series prediction model will output the baseline value of additive demand for the next 5-10 minutes.

5. The feed additive ratio control method based on environmental perception according to any one of claims 1-4, characterized in that, In step S3, the machine learning algorithm is either a fuzzy neural network or a deep reinforcement learning algorithm. If a fuzzy neural network algorithm is used, a high-quality parameter set and an environmental stress index are used as input features, and "minimizing animal stress + optimizing additive costs" is used as the output objective. A fuzzy mapping relationship between input and output is constructed to generate the optimal formulation strategy. If a deep reinforcement learning algorithm is used, "the reduction rate of the environmental stress index" and "the improvement rate of feed utilization" are used as reward functions. Through multiple rounds of iterative training, formulation decision strategies adapted to different scenarios are obtained.

6. The feed additive formulation control method based on environmental perception according to any one of claims 1-5, characterized in that, In step S5, the synchronous drive pressure supply unit generates an adaptive pressure, and establishes a positive correlation between the conveyor shaft speed of the material conveying equipment and the output pressure of the pressure supply unit through a parameter correlation model; the conveyor shaft speed is collected in real time, the target pressure value is calculated through the correlation model, and the elastic pressure component of the pressure supply unit is driven to compress the frequency, so that the output pressure matches the target pressure value, ensuring that the amount of additive sprayed is matched with the amount of feed conveyed.

7. The feed additive formulation control method based on environmental perception according to any one of claims 1-6, characterized in that, In step S6, the mixing effect parameters include additive mixing uniformity, feed nutrient retention rate, and animal short-term feed intake change rate. The specific logic of the feedback correction is as follows: if the mixing uniformity RSD value is >5%, then reduce the additive spraying flow rate by 10%-15% and increase the output pressure of the pressure supply unit by 5%-8%; if the vitamin activity retention rate is <85%, then adjust the atomized particle diameter of the spraying unit and correct the optimal ratio strategy in step S3, increasing the proportion of vitamin additives by 2%-5%.

8. An intelligent feed additive precision proportioning control system, characterized in that, The system is used to execute the feed additive ratio control method based on environmental perception as described in any one of claims 1-7, comprising: a multi-parameter acquisition module for real-time acquisition of breeding environment parameters, feed characteristic parameters, material conveying equipment operating parameters and animal physiological parameters, wherein the multi-parameter acquisition module includes a distributed sensor group; The data processing and fusion module, connected to the multi-parameter acquisition module, performs preprocessing, feature extraction, and quality assessment on the acquired multi-source parameters. The environmental stress intelligent assessment module, also connected to the data processing and fusion module, invokes a multi-factor data fusion algorithm to calculate the environmental stress index based on a high-quality parameter set and outputs environmental change trend analysis results. The intelligent prediction and decision-making module, connected to the environmental stress intelligent assessment module, incorporates a time-series prediction model and machine learning algorithm to predict additive demand baseline values ​​and generate optimal ratio strategies, while simultaneously generating control commands through an optimized control algorithm. The adaptive control execution module, connected to the intelligent prediction and decision-making module, includes an additive metering submodule, a pressure control submodule, and a spraying control submodule, precisely regulating the additive delivery rate, pressure supply unit output pressure, and spraying parameters according to control commands. The closed-loop feedback optimization module, connected to both the adaptive control execution module and the intelligent prediction and decision-making module, includes a mixing effect detection sensor and a correction algorithm unit, for real-time acquisition of mixing effect parameters and feedback of correction commands to the intelligent prediction and decision-making module to dynamically adjust the ratio strategy and control parameters.

Images