Livestock breeding environment gas dynamic monitoring system

Abstract

The invention relates to the technical field of livestock breeding environment monitoring, regulation and control, and discloses a livestock breeding environment gas dynamic monitoring system. The system comprises a data acquisition module for acquiring data by using a multi-source gas sensor; the gas component analysis module is used for analyzing the data based on a mixed spectrum decomposition algorithm to generate a gas component characteristic matrix; the environment regulation and control decision module inputs the matrix into a multi-mode collaborative decision model to generate a regulation and control instruction; the optimization model construction module is used for constructing a multi-constraint dynamic optimization model to optimize ventilation equipment parameters; and the hierarchical control execution module is used for executing the regulation and control instruction through a hierarchical architecture. The system can comprehensively monitor breeding environment gas, accurately regulate and control ventilation equipment, achieve gas concentration equalization and energy consumption minimization, improve the breeding environment quality, reduce the breeding cost and push intelligent development of the livestock breeding industry.

Description

Technical Field

[0001] This invention relates to the field of livestock farming environment monitoring and control technology, specifically a dynamic monitoring system for gases in livestock farming environment. Background Technology

[0002] In the current booming livestock farming industry, the quality of the farming environment plays a decisive role in animal health, production performance, and economic benefits. The gaseous composition of the farming environment is complex and diverse, including ammonia, hydrogen sulfide, carbon dioxide, methane, etc. Changes in the concentration of these gases not only directly affect animal growth and development, but are also closely linked to the air quality and disease transmission risk of the farming environment.

[0003] Ammonia is highly irritating; high concentrations can irritate the respiratory mucosa of animals, reduce their immunity, make them more susceptible to respiratory diseases, and affect their growth rate. Hydrogen sulfide is also highly toxic, damaging the nervous and respiratory systems of animals. Animals exposed to hydrogen sulfide for extended periods will experience a significant decline in growth performance. Excessive carbon dioxide concentrations lead to polluted air in the farming environment, making animals prone to hypoxia and thus affecting their metabolism. As a greenhouse gas, the accumulation of methane in the farming environment not only wastes energy but also indirectly reflects abnormalities in processes such as microbial fermentation, potentially indicating a deterioration of the farming environment.

[0004] However, traditional methods for monitoring the livestock farming environment have many drawbacks. Early manual testing methods are not only inefficient and time-consuming, but the results are also easily affected by human factors, making it difficult to achieve real-time and continuous monitoring of gases in the farming environment. Although some farms have introduced some simple gas monitoring equipment, these devices are relatively limited in function, usually only able to detect a few gases, and cannot comprehensively obtain gas information about the farming environment.

[0005] In terms of environmental control, existing technologies often lack precision and intelligence. Most farms employ rather crude ventilation control strategies, relying solely on experience or simple threshold settings to control ventilation equipment, without fully considering the dynamic changes in the farming environment and the interactions between different gases. This leads to either excessive ventilation causing energy waste and increased farming costs, or insufficient ventilation failing to effectively improve the farming environment, impacting animal health and production efficiency. For example, during high summer temperatures, turning on ventilation equipment based solely on temperature thresholds may overlook the rise in concentrations of harmful gases such as ammonia, thus failing to promptly ensure air quality in the farming environment.

[0006] Furthermore, different breeding areas have varying environmental characteristics and breeding needs, making it difficult for traditional uniform control methods to meet the personalized needs of breeding environment regulation. With the continuous expansion of breeding scale and the increasing refinement of breeding technology, it is urgent to develop a system that can comprehensively and in real-time monitor the gases in the livestock breeding environment and make precise and intelligent controls based on the monitoring results. This is of great significance for improving the modernization level and sustainable development capacity of the livestock breeding industry. Summary of the Invention

[0007] The purpose of this invention is to provide a dynamic monitoring system for gases in livestock farming environments to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a dynamic monitoring system for gases in livestock farming environments, the system comprising:

[0009] Data acquisition module: used to collect real-time gas concentration data in the farm through multi-source gas sensors;

[0010] Gas composition analysis module: Based on the hybrid spectral decomposition algorithm, the gas concentration data is analyzed for multi-band features to generate a gas component feature matrix;

[0011] Environmental regulation decision module: Inputs the gas component feature matrix into a pre-trained multimodal collaborative decision model to generate environmental regulation instructions;

[0012] Optimization model construction module: Constructs a multi-constraint dynamic optimization model based on the control instructions. The multi-constraint dynamic optimization model takes gas concentration equalization and energy consumption minimization as optimization objectives and uses a multi-stage decomposition optimization algorithm to globally optimize the operating parameters of the ventilation equipment.

[0013] The hierarchical control execution module outputs the optimal control strategy based on the multi-constraint dynamic optimization model and realizes the distributed execution of control commands through a hierarchical control architecture. The hierarchical control architecture includes a strategy layer, a scheduling layer, and an execution layer. The strategy layer generates a global ventilation sequence based on the control commands, the scheduling layer uses a dynamic time window prediction algorithm to adapt local ventilation parameters in real time, and the execution layer uses a fuzzy adaptive control algorithm to adjust the fan speed and ventilation angle.

[0014] Preferably, the multi-source gas sensor includes an ammonia sensor, a hydrogen sulfide sensor, a carbon dioxide sensor, a methane sensor, and a temperature and humidity sensor; the acquisition of real-time gas concentration data in the farm through the multi-source gas sensor includes:

[0015] Spatiotemporal alignment of ammonia and hydrogen sulfide sensor data was performed to construct a three-dimensional gas diffusion distribution map; multi-scale wavelet decomposition of carbon dioxide and methane sensor data was performed to generate gas concentration change feature sequences.

[0016] A dual-branch feature fusion network is constructed. The first branch uses a three-dimensional convolutional network to extract the spatial distribution features of the three-dimensional gas diffusion distribution map, and the second branch uses a gated recurrent unit to extract the temporal correlation features of the gas concentration change feature sequence.

[0017] The spatial distribution features and temporal correlation features are fused through a cross-modal attention mechanism to generate a joint feature tensor. Based on an adaptive weighting mechanism, dynamic feature selection is performed on the joint feature tensor to output comprehensive monitoring features including gas distribution, ambient temperature and humidity, and equipment status.

[0018] Preferably, the multimodal collaborative decision-making model adopts a hierarchical graph attention network structure and generates control instructions based on a dynamic feature allocation mechanism; the hierarchical graph attention network structure includes:

[0019] Construct a farm-equipment interaction diagram. The nodes in the diagram include ventilation equipment nodes, gas source nodes, environmental nodes, and sensor nodes. The node attributes include gas concentration gradient, equipment operating temperature, and ventilation efficiency.

[0020] A two-stage graph attention mechanism is adopted. In the first stage, the interaction weights between ventilation equipment nodes and adjacent nodes are calculated through the spatial graph attention layer. In the second stage, the importance of historical gas concentration changes is screened through the time series graph attention layer.

[0021] The node attributes are iteratively updated based on the multi-head graph attention module, and each attention head integrates the node attributes and external environment parameters; the training process is stabilized through residual connections and layer normalization mechanisms, and the final output includes control instructions containing device constraints and environmental thresholds.

[0022] Preferably, the multi-stage decomposition optimization algorithm integrates dynamic constraint relaxation and weight adaptive strategies, including:

[0023] The ventilation parameter optimization problem is modeled as a multi-constraint mixed integer nonlinear programming problem, with decision variables including discrete equipment start-up and shutdown variables and continuous speed adjustment variables;

[0024] The relaxation problem is initialized and the Pareto front approximation solution is calculated. A dynamic weight allocation mechanism is used to update the objective function weight coefficients according to the real-time environment requirements.

[0025] In the decomposition phase, the global problem is divided into multiple single-constraint subproblems based on the constraint space decomposition technique; in the coordination phase, a relaxation factor is introduced to balance the conflicting constraints between the subproblems.

[0026] A parallel solver is used to iteratively optimize the subproblem. In each iteration, the local solution is updated and the global solution is synchronized through a coordination factor.

[0027] Preferably, the dynamic time window prediction algorithm adapts local ventilation parameters in real time, including:

[0028] A time-varying environment prediction model is constructed, and the gas diffusion dynamics equation is discretized into a state-space model. The state-space model includes gas concentration prediction error, wind speed disturbance term, and ventilation angle differential term.

[0029] The objective function for sliding window optimization is designed, which includes gas concentration fluctuation suppression, equipment energy consumption balancing, and environmental temperature and humidity compensation.

[0030] Preferably, the execution layer uses a fuzzy adaptive control algorithm to adjust the fan speed and ventilation angle, including:

[0031] Design a two-layer fuzzy inference system that maps rotation speed error and ventilation angle error to fuzzy rule activation degree; construct an adaptive membership function adjustment mechanism to dynamically adjust the coverage of the fuzzy set according to the error change rate.

[0032] Preferably, the three-dimensional convolutional network employs a multi-scale pyramid structure to accelerate feature extraction, including:

[0033] The three-dimensional gas diffusion distribution map is divided into a multi-resolution cubic grid, and each grid cell stores the gas concentration gradient and diffusion rate statistics.

[0034] In the encoding stage, dilated convolutional kernels are used to expand the spatial receptive field, and in the decoding stage, deconvolutional layers are used to restore the three-dimensional spatial details.

[0035] A spatial attention module is introduced to reweight the channels of the feature map.

[0036] Preferably, the spatial graph attention layer employs a relative attribute encoding mechanism, including:

[0037] Define the relative attribute vectors between ventilation equipment nodes and adjacent nodes, including gas diffusion distance, ventilation efficiency coefficient, and temperature and humidity differences;

[0038] The relative attribute vector is mapped to the bias parameters of the graph attention kernel through a nonlinear transformation layer;

[0039] The bias parameters are superimposed on the standard graph attention operation.

[0040] Preferably, the dynamic weight allocation mechanism is implemented based on an online reinforcement learning strategy, including:

[0041] Collect historical optimization process data of objective function values ​​and constraint violation degrees as training samples, and construct a radial basis function network to fit the mapping relationship between weight coefficients and environmental state;

[0042] The network parameters are updated online using incremental gradient descent, and the weight distribution of the objective function is adjusted in real time.

[0043] When a sudden change in gas concentration is detected, an emergency weight reset operation is triggered.

[0044] Preferably, the time-varying environment prediction model is implemented through orthogonal decomposition, including:

[0045] The uncertainty of gas concentration is modeled as an interval variable, with the upper and lower bounds determined by the standard deviation of the prediction error. The interval variable is orthogonally decomposed to separate the deterministic and random components.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] The livestock farming environment gas dynamic monitoring system of this invention has many significant benefits. In terms of monitoring, the data acquisition module, through various gas sensors, can comprehensively acquire data on ammonia, hydrogen sulfide, carbon dioxide, methane, as well as temperature and humidity, providing richer information compared to traditional monitoring equipment. Its spatiotemporal alignment, wavelet decomposition, and feature fusion operations deeply mine the value of the data, generating comprehensive monitoring features that accurately reflect the farming environment, providing a reliable basis for subsequent regulation.

[0048] The gas composition analysis module uses a hybrid spectral decomposition algorithm to perform multi-band feature analysis on gas concentration data, which can accurately identify complex gas components. The generated gas component feature matrix provides key support for environmental regulation decisions and helps to accurately grasp environmental change trends.

[0049] The environmental control decision-making module employs a pre-trained multimodal collaborative decision-making model, fully considering various factors such as equipment and environment. Its hierarchical graph attention network structure, through a dynamic feature allocation mechanism, generates scientifically sound and reasonable control instructions, enabling precise decisions for different aquaculture scenarios and avoiding the blindness of traditional experience-based control.

[0050] The optimization model construction module constructs a multi-constraint dynamic optimization model with the goals of achieving gas concentration equalization and minimizing energy consumption. A multi-stage decomposition optimization algorithm is then used to globally optimize the ventilation equipment parameters. This optimization method significantly reduces ventilation energy consumption while ensuring suitable gas conditions in the aquaculture environment. For example, by rationally adjusting the start-up, shutdown, and speed of the ventilation equipment, unnecessary energy consumption is reduced, aquaculture costs are lowered, and efficient resource utilization is achieved.

[0051] The hierarchical architecture of the hierarchical control execution module has a clear division of labor. The strategy layer generates a global ventilation sequence to plan the ventilation scheme from a macro perspective; the scheduling layer uses a dynamic time window prediction algorithm to adapt local ventilation parameters in real time to adapt to dynamic changes in the environment; the execution layer uses a fuzzy adaptive control algorithm to precisely adjust the fan speed and ventilation angle to achieve efficient execution of control commands and ensure that the aquaculture environment is always in a stable and suitable state.

[0052] Overall, the system enables dynamic monitoring and precise control of gases in the livestock farming environment, significantly improving the quality of the farming environment, reducing the harm of harmful gases to animal health, lowering disease incidence, and improving animal growth performance and economic benefits. At the same time, the system's intelligent and automated operation reduces human intervention, improves farming management efficiency, and promotes the modernization and intelligentization of the livestock farming industry. Attached Figure Description

[0053] Figure 1 This is a schematic diagram illustrating the working principle of the livestock farming environment gas dynamic monitoring system described in this invention.

[0054] Figure 2 A detailed schematic diagram of the data acquisition module;

[0055] Figure 3 This is a schematic diagram illustrating the working principle of a multimodal collaborative decision-making model.

[0056] Figure 4 This is a diagram illustrating the working principle of the dynamic time window prediction algorithm. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] Please see Figures 1-4 This invention provides a dynamic monitoring system for gaseous environments in livestock farming, and its specific implementation is described in detail below.

[0059] Data Acquisition Module: This module utilizes multi-source gas sensors to collect real-time gas concentration data within the farm. These sensors include various types such as ammonia sensors, hydrogen sulfide sensors, carbon dioxide sensors, methane sensors, and temperature and humidity sensors. Through their collaborative operation, comprehensive information on the concentration of various gases within the farm, as well as environmental temperature and humidity data, can be obtained, providing fundamental data support for subsequent analysis and decision-making.

[0060] Gas Composition Analysis Module: This module, based on a hybrid spectral decomposition algorithm, performs multi-band feature analysis on the gas concentration data acquired by the data acquisition module. This analytical method extracts the characteristics of different gas components from complex gas concentration data, thereby generating a gas component feature matrix. This matrix records detailed characteristic information of various gas components and serves as a crucial basis for subsequent environmental control decisions.

[0061] Environmental control decision-making module: The gas component feature matrix generated by the gas composition analysis module is input into the pre-trained multimodal collaborative decision-making model. Based on the input feature matrix and its learned knowledge and patterns, the model generates corresponding environmental control instructions. These instructions specify the control measures required for the current gas conditions in the aquaculture environment to ensure its suitability.

[0062] The optimization model construction module constructs a multi-constraint dynamic optimization model based on the control instructions generated by the environmental control decision module. This model takes gas concentration equalization and energy consumption minimization as optimization objectives, and uses a multi-stage decomposition optimization algorithm to globally optimize the operating parameters of the ventilation equipment. Through this optimization, the energy consumption of the ventilation equipment can be minimized while meeting the gas concentration control requirements, achieving efficient resource utilization.

[0063] The hierarchical control execution module, based on the optimal control strategy output by a multi-constraint dynamic optimization model, implements distributed execution of control commands through a hierarchical control architecture. This architecture comprises three layers: a strategy layer, a scheduling layer, and an execution layer. The strategy layer generates a global ventilation sequence based on the control commands, macroscopically planning the overall ventilation schedule. The scheduling layer uses a dynamic time window prediction algorithm to adapt local ventilation parameters in real time to accommodate dynamic changes in the aquaculture environment. The execution layer uses a fuzzy adaptive control algorithm to adjust fan speed and ventilation angle, precisely controlling the operation of ventilation equipment and thus achieving effective control of gases in the aquaculture environment.

[0064] The technical solution of the present invention will be further described in detail below with reference to specific embodiments.

[0065] Example 1:

[0066] In this embodiment, the specific process of data acquisition by the multi-source gas sensor in the data acquisition module is described in detail.

[0067] The multi-source gas sensors include ammonia, hydrogen sulfide, carbon dioxide, methane, and temperature and humidity sensors. When collecting real-time gas concentration data within the farm, ammonia and hydrogen sulfide sensor data are spatiotemporally aligned to construct a three-dimensional gas diffusion distribution map. This is because ammonia and hydrogen sulfide are harmful gases that significantly impact animal health and growth in livestock farming environments, making their spatial and temporal distribution crucial. Spatiotemporal alignment integrates ammonia and hydrogen sulfide sensor data collected at different locations and times, constructing a three-dimensional gas diffusion distribution map that visually displays the diffusion trends of these two gases within the farm.

[0068] Multi-scale wavelet decomposition was performed on carbon dioxide and methane sensor data to generate gas concentration change characteristic sequences. Carbon dioxide and methane are important indicators reflecting air quality and animal metabolism in aquaculture environments. Multi-scale wavelet decomposition can analyze gas concentration changes at different time scales, capturing subtle changes and trends in the data, thereby generating more representative gas concentration change characteristic sequences.

[0069] A dual-branch feature fusion network is constructed. The first branch uses a 3D convolutional network to extract the spatial distribution features of the 3D gas diffusion distribution map. The 3D convolutional network can perform convolution operations on 3D data, and by sliding the convolution kernel in 3D space, it extracts spatial features at different locations in the gas diffusion distribution map, such as areas with high gas concentration and diffusion boundaries. The second branch uses a gated recurrent unit to extract the temporal correlation features of the gas concentration change sequence. The gated recurrent unit can effectively process time-series data, controlling the flow of information through a gating mechanism to better capture the changing patterns of gas concentration over time and the correlation information between different time points.

[0070] A joint feature tensor is generated by fusing spatial distribution features and temporal correlation features through a cross-modal attention mechanism. This mechanism weights the fusion based on the importance of different features, ensuring the resulting joint feature tensor includes both spatial gas distribution information and temporal correlation information. An adaptive weighting mechanism dynamically selects features from the joint feature tensor, outputting comprehensive monitoring features encompassing gas distribution, ambient temperature and humidity, and equipment status. This adaptive weighting mechanism automatically adjusts the weights of different features based on the actual conditions of the current aquaculture environment, highlighting features more relevant to the current environment. This results in comprehensive and accurate monitoring features, providing high-quality data support for subsequent gas composition analysis and environmental control.

[0071] Example 2:

[0072] This embodiment details the working process of the multimodal collaborative decision-making model.

[0073] The multimodal collaborative decision-making model employs a hierarchical graph attention network structure and generates control commands based on a dynamic feature allocation mechanism. First, a farm-equipment interaction graph is constructed. Nodes in the graph include ventilation equipment nodes, gas source nodes, environmental nodes, and sensor nodes. Node attributes include gas concentration gradient, equipment operating temperature, and ventilation efficiency. Ventilation equipment nodes represent various ventilation devices within the farm, and their ventilation efficiency directly affects the ventilation effect. Gas source nodes represent the sources of gas generation, such as animal activity areas, and gas concentration gradients reflect the direction and speed of gas diffusion. Environmental nodes encompass factors such as temperature and humidity in the farming environment. Sensor nodes correspond to various gas sensors and temperature and humidity sensors, and their attributes record real-time monitoring data.

[0074] A two-stage graph attention mechanism is employed. The first stage calculates the interaction weights between ventilation equipment nodes and their neighboring nodes through a spatial graph attention layer. This spatial graph attention layer uses a relative attribute encoding mechanism to define relative attribute vectors between ventilation equipment nodes and their neighboring nodes, including gas diffusion distance, ventilation efficiency coefficient, and temperature and humidity differences. Gas diffusion distance represents the distance between the ventilation equipment and the gas source or other areas affecting ventilation; the closer the distance, the greater the impact on gas diffusion. The ventilation efficiency coefficient reflects the contribution of the ventilation equipment's own performance to the overall ventilation effect. Temperature and humidity differences reflect the difference in temperature and humidity between the environment surrounding the ventilation equipment and other areas, which affects gas diffusion and ventilation efficiency. A nonlinear transformation layer maps the relative attribute vectors to bias parameters of the graph attention kernel, and then the bias parameters are superimposed on the standard graph attention operation. This allows for more accurate calculation of the spatial interaction weights between ventilation equipment nodes and their neighboring nodes, highlighting node information closely related to the ventilation equipment.

[0075] The second stage uses a time-series graph attention layer to filter historical gas concentration changes based on their importance. Historical gas concentration changes record the evolution of gas conditions in the aquaculture environment over time. The time-series graph attention layer can filter these historical data based on their importance, focusing on historical moments that have a significant impact on current environmental control decisions while ignoring less important fluctuations, making decisions more targeted and stable.

[0076] The multi-head graph attention module iteratively updates node attributes, with each attention head fusing node attributes with external environmental parameters. This module analyzes and fuses node attributes from different perspectives, incorporating external environmental parameters such as weather changes and variations in the number of farmed animals to provide a more comprehensive update. Residual connections and layer normalization mechanisms stabilize the training process, preventing gradient vanishing or exploding issues and ensuring stable model convergence. The final output includes control instructions containing equipment constraints and environmental thresholds. These instructions clearly define the operational limitations of ventilation equipment and the appropriate ranges for maintaining parameters such as gas concentration, temperature, and humidity in the farming environment, providing clear guidance for subsequent environmental control.

[0077] Example 3:

[0078] This embodiment explores in detail the specific implementation of the multi-stage decomposition optimization algorithm.

[0079] A multi-stage decomposition optimization algorithm integrates dynamic constraint relaxation and weight adaptive strategies. First, the ventilation parameter optimization problem is modeled as a multi-constraint mixed-integer nonlinear programming problem, with decision variables including discrete equipment start-up / shutdown variables and continuous speed adjustment variables. Let the equipment start-up / shutdown variables be... ( , The number of ventilation equipment, , This indicates that the equipment has stopped. (Indicates device startup), speed adjustment variable is ( , The number of devices with adjustable speed. The value range is determined based on the actual speed adjustment range of the equipment. The objective function of this planning problem is... It includes two objectives: equalizing gas concentration and minimizing energy consumption. The constraints include various limitations such as the physical limitations of the equipment and the safe range of gas concentration.

[0080] The relaxation problem is initialized, and the Pareto front approximation solution is calculated. The Pareto front approximation solution is a set of optimal solutions found among multiple conflicting objectives, such that the performance of a particular objective cannot be further improved without degrading the performance of other objectives. A dynamic weight allocation mechanism is employed to update the objective function weight coefficients according to real-time environmental requirements. This dynamic weight allocation mechanism is implemented based on an online reinforcement learning strategy, collecting historical objective function values ​​and constraint violation degrees as training samples to construct a radial basis function network that fits the mapping relationship between the weight coefficients and the environmental state. Let the objective function weight coefficients be... ( (These correspond to the objectives of gas concentration equalization and energy consumption minimization, respectively), and the environmental state is... Through radial basis function network To fit and The relationship, that is The network parameters are updated online using incremental gradient descent, and the objective function weights are adjusted in real time, allowing the weights to adaptively adjust according to dynamic changes in the aquaculture environment. When a sudden change in gas concentration is detected, an emergency weight reset operation is triggered to ensure rapid adjustment of the optimization direction under abnormal conditions and to safeguard the safety of the aquaculture environment.

[0081] In the decomposition phase, the global problem is divided into multiple single-constraint subproblems based on constraint space decomposition techniques. This simplifies complex multi-constraint problems and facilitates their separate solutions. In the coordination phase, relaxation factors are introduced to balance conflicting constraints between subproblems. ( This is used to adjust the relationships between different subproblems, avoiding conflicts between subproblems that could lead to unreasonable optimization results. A parallel solver is used to iteratively optimize the subproblems, updating the local solution in each iteration and synchronizing the global solution through a coordination factor. Coordination factor ( This is used to control the degree of influence of local solutions on global solutions, ensuring that while optimizing locally, the consistency and optimality of global solutions can be maintained. Through continuous iteration, the global optimal solution of the ventilation equipment operating parameters is finally obtained, achieving the goals of equalizing gas concentration and minimizing energy consumption.

[0082] Example 4:

[0083] In this embodiment, the process of adapting local ventilation parameters in real time using a dynamic time window prediction algorithm will be described in detail.

[0084] Discretizing the gas diffusion kinetics equation into a state-space model is a crucial step in constructing a time-varying environmental prediction model. The gas diffusion kinetics equation describes the diffusion patterns of gases in the aquaculture environment, and its expression is: .in, This represents the concentration of a specific gas in the farming environment. Different gas concentrations have varying degrees of impact on the health and growth of farmed animals. It indicates time, because the gas conditions in the breeding environment change constantly over time, so time is an important variable; These are spatial coordinates used to determine the specific location of the gas within the farm; the gas concentration may vary significantly at different locations. It is the gas diffusion coefficient, which depends on the type of gas and the physical characteristics of the breeding environment. For example, factors such as air flow, temperature and humidity will affect the size of the gas diffusion coefficient. The term "source" represents the source of gas production. In livestock farms, animal respiration and the decomposition of excrement are common sources of gas production.

[0085] After discretizing the equation into a state-space model, the model includes gas concentration prediction errors. Wind speed disturbance item and ventilation angle differential term Gas concentration prediction error Used to measure the deviation between predicted and actual gas concentrations, it reflects the accuracy of the prediction model. Wind speed disturbance term. This reflects the difference between actual and ideal wind speeds. In actual aquaculture environments, changes in external natural wind and unstable operation of ventilation equipment can cause wind speed disturbances, which in turn affect the diffusion and distribution of gases. (Ventilation angle differential term) This indicates a minute change in the ventilation angle. Adjusting the ventilation angle plays an important role in the flow and diffusion path of gases. Even a small change in angle can cause changes in the distribution of gases within the farm.

[0086] When designing the sliding window optimization objective function, gas concentration fluctuation suppression, equipment energy consumption balancing, and environmental temperature and humidity compensation terms were comprehensively considered. The length of the sliding window was set to... Within this window, the gas concentration at different times is used... ( This indicates that the equipment's energy consumption... The deviations between the ambient temperature and humidity and the set values ​​are represented by... and express.

[0087] Gas concentration fluctuation suppression term This value measures the degree of fluctuation in gas concentration within a sliding window. A smaller value indicates that the gas concentration remains relatively stable during this period, which is beneficial for maintaining a stable farming environment. Conversely, a larger value indicates drastic fluctuations in gas concentration, which may adversely affect the health of farmed animals. For example, large fluctuations in ammonia concentration may irritate the respiratory tract of animals and cause disease.

[0088] Equipment energy consumption balancing item ,in This item's function is to ensure that the equipment's energy consumption remains balanced over a period of time. In actual operation, frequent start-ups and shutdowns or significant power fluctuations of ventilation equipment can lead to unstable energy consumption, increasing operating costs and potentially affecting the equipment's lifespan. By optimizing this item, ventilation equipment can meet ventilation needs while minimizing energy consumption fluctuations.

[0089] Environmental temperature and humidity compensation item , and These are the temperature and humidity compensation coefficients. These two coefficients are determined based on the sensitivity of the actual breeding environment to temperature and humidity. Different types of farmed animals have different requirements for temperature and humidity; for example, chicks are more sensitive to changes in temperature and humidity. When the environmental temperature and humidity deviate from the set values... and When the temperature is high, compensation is made by adjusting ventilation parameters to maintain a suitable temperature and humidity environment.

[0090] By optimizing the objective function through this sliding window, local ventilation parameters can be adapted in real time according to the actual conditions of the current breeding environment. For example, when large fluctuations in gas concentration are detected, the ventilation volume can be appropriately increased to stabilize the gas concentration; when equipment energy consumption is too high, the operating mode of the ventilation equipment can be adjusted to reduce energy consumption; when temperature and humidity deviate from the set values, the ventilation angle and wind speed can be adjusted accordingly. This allows the ventilation system to meet the requirements for gas concentration control while also ensuring the stability of equipment energy consumption and ambient temperature and humidity, thus creating a good growth environment for the farmed animals.

[0091] Example 5:

[0092] This embodiment will detail the specific method by which the execution layer implements the adjustment of fan speed and ventilation angle based on the fuzzy adaptive control algorithm.

[0093] When designing a two-layer fuzzy inference system, the rotational speed error... Error with ventilation angle This is an important input parameter. Speed ​​error. Based on the actual speed of the fan With target speed Subtraction yields, i.e. Target speed It is preset according to the needs of the breeding environment. For example, when the ammonia concentration in the breeding farm is high, it is necessary to increase the fan speed to increase the ventilation volume and reduce the ammonia concentration; while the actual speed This is obtained through real-time measurement using sensors. Ventilation angle error. The actual ventilation angle of the fan ventilation angle relative to the target The difference, i.e. Target ventilation angle The setting depends on the layout of the farm and the gas distribution. For example, in areas where the breeding areas are concentrated, the ventilation angle needs to be adjusted to better cover these areas; the actual ventilation angle... It is also monitored in real time by sensors.

[0094] First of all and Fuzzification is performed, transforming the error into fuzzy linguistic variables such as "positive large," "positive small," "zero," "negative small," and "negative large." This is because actual error values ​​are continuous numerical values, while fuzzy control needs to convert them into a more easily understood and processed linguistic form. For example, when the speed error... When the value is large and positive, it is fuzzified to "positive and large", indicating that the actual fan speed is much higher than the target speed; when the ventilation angle error When the value is small and negative, it is blurred into "negative small", indicating that the actual ventilation angle is slightly smaller than the target ventilation angle.

[0095] Based on a pre-defined fuzzy rule base, the activation degree of fuzzy rules is calculated through fuzzy inference. The rules in the fuzzy rule base are summarized from a large amount of practical experience and experimental data. For example, the rule "If the fan speed error is positive and the ventilation angle error is positive and small, then increase the fan speed and appropriately decrease the ventilation angle" means that when the fan speed is too high and the ventilation angle is slightly too large, in order to better control the gas distribution in the aquaculture environment, it is necessary to further increase the fan speed to enhance the ventilation effect, while appropriately decreasing the ventilation angle to concentrate the ventilation more in the required area. In the actual inference process, based on the input fuzzified fan speed error and ventilation angle error, matching rules are searched in the fuzzy rule base, and the activation degree of each rule is calculated. The activation degree reflects the applicability of the rule under the current circumstances.

[0096] An adaptive membership function adjustment mechanism is constructed to dynamically adjust the coverage of the fuzzy set based on the error change rate. (Speed ​​error change rate) and ventilation angle error change rate It is an important indicator for measuring the dynamic changes of a system. When the error rate of change is large, it means that the system state is changing rapidly, and in this case, it is necessary to expand the coverage of the fuzzy set. For example, if the speed error rate of change is large... A sudden increase indicates a rapid change in fan speed. To adjust the fan speed more flexibly, the actual error range corresponding to linguistic variables such as "positive large" and "positive small" in the fuzzy set needs to be expanded to more comprehensively control such rapid changes. Conversely, when the error rate of change is small, reducing the coverage of the fuzzy set can improve control accuracy. For example, the rate of change of ventilation angle error... When the value is relatively small, it indicates that the change in the ventilation angle is relatively stable. In this case, reducing the coverage of the fuzzy set can more accurately adjust the ventilation angle and make it closer to the target value.

[0097] Through this adaptive adjustment mechanism, the fuzzy adaptive control algorithm can better adapt to the dynamic changes in the breeding environment. In actual operation, the breeding environment is affected by various factors, such as animal activity and weather changes, which cause parameters such as gas concentration, temperature, and humidity to change continuously. The execution layer based on the fuzzy adaptive control algorithm can adjust the fan speed and ventilation angle in real time according to these changes, ensuring that the ventilation system can provide appropriate ventilation volume and direction according to actual needs, maintaining the stability and suitability of the gas environment in the breeding environment, and ensuring the healthy growth of the farmed animals.

[0098] Example 6:

[0099] This embodiment describes the specific process of using a multi-scale pyramid structure in a 3D convolutional network to accelerate feature extraction.

[0100] The 3D gas diffusion map is divided into a multi-resolution cubic mesh, with each mesh cell storing the gas concentration gradient and diffusion rate statistics. The multi-resolution cubic mesh allows for analysis of the gas diffusion map at different scales; large-scale meshes capture the overall gas diffusion trend, while small-scale meshes focus on local details. The gas concentration gradient reflects the rate and direction of change in gas concentration in space, and the diffusion rate statistics record information such as the average velocity of gas diffusion. These data provide rich information for subsequent feature extraction.

[0101] In the encoding stage, dilated convolutional kernels are used to expand the spatial receptive field. During the convolution operation, there is a certain spacing between the elements of the dilated convolutional kernel. This expands the receptive field of the kernel without increasing its size, thereby capturing a wider range of spatial information. Let the dilation rate of the dilated convolutional kernel be... The size of a regular convolution kernel is Therefore, the actual receptive field size of the expanded convolution kernel will be larger than the calculated value. It can capture gas concentration information at greater distances.

[0102] The decoding stage uses deconvolutional layers to recover 3D spatial details. Deconvolutional layers upsample the feature maps extracted in the encoding stage to recover the spatial details of the 3D gas diffusion distribution map, making the resolution of the feature maps similar to the original image, which facilitates subsequent analysis and processing.

[0103] A spatial attention module is introduced to reweight the channels of the feature map. The spatial attention module can weight channels based on their importance in the feature map. Let the feature map be... The number of channels is The spatial attention module calculates the weights for each channel. ( The feature map is weighted by each channel to obtain the weighted feature map. This approach highlights channels that are more important for expressing gas diffusion characteristics, suppresses information from irrelevant channels, and improves the accuracy and effectiveness of feature extraction, thereby better supporting subsequent gas composition analysis and environmental control decisions.

[0104] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0105] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A dynamic monitoring system for gases in livestock farming environments, characterized in that, include: Data acquisition module: used to collect real-time gas concentration data in the farm through multi-source gas sensors; Gas composition analysis module: Based on the hybrid spectral decomposition algorithm, the gas concentration data is analyzed for multi-band features to generate a gas component feature matrix; Environmental regulation decision module: Inputs the gas component feature matrix into a pre-trained multimodal collaborative decision model to generate environmental regulation instructions; Optimization model construction module: Constructs a multi-constraint dynamic optimization model based on the control instructions. The multi-constraint dynamic optimization model takes gas concentration equalization and energy consumption minimization as optimization objectives and uses a multi-stage decomposition optimization algorithm to globally optimize the operating parameters of the ventilation equipment. The hierarchical control execution module outputs the optimal control strategy based on the multi-constraint dynamic optimization model and realizes the distributed execution of control commands through a hierarchical control architecture. The hierarchical control architecture includes a strategy layer, a scheduling layer, and an execution layer. The strategy layer generates a global ventilation sequence based on the control commands, the scheduling layer uses a dynamic time window prediction algorithm to adapt local ventilation parameters in real time, and the execution layer uses a fuzzy adaptive control algorithm to adjust the fan speed and ventilation angle.

2. The livestock farming environment gas dynamic monitoring system according to claim 1, characterized in that, The multi-source gas sensor includes an ammonia sensor, a hydrogen sulfide sensor, a carbon dioxide sensor, a methane sensor, and a temperature and humidity sensor; the real-time gas concentration data collection in the farm via the multi-source gas sensor includes: Spatiotemporal alignment of ammonia and hydrogen sulfide sensor data was performed to construct a three-dimensional gas diffusion distribution map; multi-scale wavelet decomposition of carbon dioxide and methane sensor data was performed to generate gas concentration change feature sequences. A dual-branch feature fusion network is constructed. The first branch uses a three-dimensional convolutional network to extract the spatial distribution features of the three-dimensional gas diffusion distribution map, and the second branch uses a gated recurrent unit to extract the temporal correlation features of the gas concentration change feature sequence. The spatial distribution features and temporal correlation features are fused through a cross-modal attention mechanism to generate a joint feature tensor. Based on an adaptive weighting mechanism, dynamic feature selection is performed on the joint feature tensor to output comprehensive monitoring features including gas distribution, ambient temperature and humidity, and equipment status.

3. The livestock farming environment gas dynamic monitoring system according to claim 1, characterized in that, The multimodal collaborative decision-making model employs a hierarchical graph attention network structure and generates control instructions based on a dynamic feature allocation mechanism; the hierarchical graph attention network structure includes: Construct a farm-equipment interaction diagram. The nodes in the diagram include ventilation equipment nodes, gas source nodes, environmental nodes, and sensor nodes. The node attributes include gas concentration gradient, equipment operating temperature, and ventilation efficiency. A two-stage graph attention mechanism is adopted. In the first stage, the interaction weights between ventilation equipment nodes and adjacent nodes are calculated through the spatial graph attention layer. In the second stage, the importance of historical gas concentration changes is screened through the time series graph attention layer. The node attributes are iteratively updated based on the multi-head graph attention module, and each attention head integrates the node attributes and external environment parameters; the training process is stabilized through residual connections and layer normalization mechanisms, and the final output includes control instructions containing device constraints and environmental thresholds.

4. The livestock farming environment gas dynamic monitoring system according to claim 1, characterized in that, The multi-stage decomposition optimization algorithm integrates dynamic constraint relaxation and weight adaptive strategies, including: The ventilation parameter optimization problem is modeled as a multi-constraint mixed integer nonlinear programming problem, with decision variables including discrete equipment start-up and shutdown variables and continuous speed adjustment variables; The relaxation problem is initialized and the Pareto front approximation solution is calculated. A dynamic weight allocation mechanism is used to update the objective function weight coefficients according to the real-time environment requirements. In the decomposition phase, the global problem is divided into multiple single-constraint subproblems based on the constraint space decomposition technique; in the coordination phase, a relaxation factor is introduced to balance the conflicting constraints between the subproblems. A parallel solver is used to iteratively optimize the subproblem. In each iteration, the local solution is updated and the global solution is synchronized through a coordination factor.

5. The livestock farming environment gas dynamic monitoring system according to claim 1, characterized in that, The dynamic time window prediction algorithm adapts local ventilation parameters in real time, including: A time-varying environment prediction model is constructed, and the gas diffusion dynamics equation is discretized into a state-space model. The state-space model includes gas concentration prediction error, wind speed disturbance term, and ventilation angle differential term. The objective function for sliding window optimization is designed, which includes gas concentration fluctuation suppression, equipment energy consumption balancing, and environmental temperature and humidity compensation.

6. The livestock farming environment gas dynamic monitoring system according to claim 1, characterized in that, The execution layer uses a fuzzy adaptive control algorithm to adjust the fan speed and ventilation angle, including: Design a two-layer fuzzy inference system that maps rotation speed error and ventilation angle error to fuzzy rule activation degree; construct an adaptive membership function adjustment mechanism to dynamically adjust the coverage of the fuzzy set according to the error change rate.

7. The livestock farming environment gas dynamic monitoring system according to claim 2, characterized in that, The three-dimensional convolutional network employs a multi-scale pyramid structure to accelerate feature extraction, including: The three-dimensional gas diffusion distribution map is divided into a multi-resolution cubic grid, and each grid cell stores the gas concentration gradient and diffusion rate statistics. In the encoding stage, dilated convolutional kernels are used to expand the spatial receptive field, and in the decoding stage, deconvolutional layers are used to restore the three-dimensional spatial details. A spatial attention module is introduced to reweight the channels of the feature map.

8. The livestock farming environment gas dynamic monitoring system according to claim 3, characterized in that, The spatial graph attention layer employs a relative attribute encoding mechanism, including: Define the relative attribute vector between the ventilation equipment node and its adjacent nodes, including gas diffusion distance, ventilation efficiency coefficient, and temperature and humidity difference; The relative attribute vector is mapped to the bias parameters of the graph attention kernel through a nonlinear transformation layer; The bias parameters are superimposed on the standard graph attention operation.

9. The livestock farming environment gas dynamic monitoring system according to claim 4, characterized in that, The dynamic weight allocation mechanism is implemented based on an online reinforcement learning strategy, including: Collect historical optimization process data of objective function values ​​and constraint violation degrees as training samples, and construct a radial basis function network to fit the mapping relationship between weight coefficients and environmental state; The network parameters are updated online using incremental gradient descent, and the weight distribution of the objective function is adjusted in real time. When a sudden change in gas concentration is detected, an emergency weight reset operation is triggered.

10. The livestock farming environment gas dynamic monitoring system according to claim 5, characterized in that, The time-varying environment prediction model is implemented through orthogonal decomposition, including: The uncertainty of gas concentration is modeled as an interval variable, with the upper and lower bounds determined by the standard deviation of the prediction error. The interval variable is orthogonally decomposed to separate the deterministic and random components.

Images