Disclosure of Invention
The invention aims to design a self-adaptive intelligent control method and a self-adaptive intelligent control system for poultry breeding environments, which not only need to have the capability of sensing the environments and the poultry states in real time, but also can dynamically adjust the environment parameters according to actual requirements, so that the energy consumption is reduced and the overall benefit of breeding is improved while the health and the production efficiency of the poultry are optimized.
In order to achieve the above object, in a first aspect of the present invention, there is provided an adaptive intelligent control method for poultry farming environments, the method comprising:
s1, installing a miniature sensor on an individual poultry body, collecting biological signals of the individual poultry body, preprocessing the collected data, and then designing a multidimensional dynamic boundary abnormality detection algorithm to correct abnormal points to obtain dataData are subjected to design dynamic feature weighting extraction methodExtracting weighted characteristics including heart rate average valueAverage value of body temperatureAnd activity weighted variance
S2, constructing a health state model based on a health index by using the extracted characteristics, judging the past state of the poultry by calculating the health index Hi (t) to obtain the deviation degree delta Hi (t) of the estimated health state, then designing a feedback weight self-adaptive adjustment algorithm to carry out self-adaptive adjustment on the weight in the health state model, classifying the health state of the poultry based on the health index Hi (t) and the deviation delta Hi (t), and adopting a corresponding environment regulation strategy;
Wherein, the past state of the poultry is judged by calculating the health index Hi (t) to obtain the deviation degree Δhi (t) of the evaluation health state:
if ΔHi (t) exceeds a preset dynamic threshold ΘH, indicating that the bird was in an abnormal or stressed state in the past;
S3, acquiring position coordinate data and speed vector data of the poultry, mapping the position coordinate data and the speed vector data to behavior feature vectors Bj (t) in a feature space, identifying a behavior pattern of a poultry group according to the behavior feature vectors Bj (t) by adopting an adaptive clustering algorithm based on the behavior feature vectors, serializing according to the behavior pattern, and constructing a time sequence model based on the behavior pattern to monitor the change of the behavior pattern of the group in real time;
S4, designing a multi-mode data fusion and weighted decision model to fuse the health index Hi (t), the behavior feature vector Bj (t) and the environmental sensor data to obtain a multi-mode feature vector F (t), generating a comprehensive environmental control index D (t) through a self-adaptive weighting function, dynamically adjusting the environmental control strategy according to the comprehensive environmental control index D (t), and carrying out real-time feedback monitoring after adjustment;
S5, constructing a composite growth index Gi (t) for identifying a growth stage according to a biological signal, designing a dynamic environment demand vector E(g) (t) according to the composite growth index Gi (t), then designing a growth stage differentiation weight distribution algorithm to dynamically adjust the priority weight of the dynamic environment demand vector E(g) (t), verifying through a specific growth stage feedback mechanism after each regulation, and finally designing a comprehensive optimization mechanism of multi-stage environment control to integrate data of different growth stages;
s6, designing a self-adaptive multi-objective optimization model to dynamically adjust a control strategy according to real-time data, then designing an individual difference self-adaptive control model to distribute individualized control parameters thetai (t) for each individual according to biological signals, dynamically adjusting the control parameters by combining historical data, and designing a multi-level feedback loop and emergency response mechanism to ensure timely response of the system under emergency conditions.
In one embodiment, the multidimensional dynamic boundary anomaly detection algorithm specifically includes:
based on the local mean μi (t) and standard deviation σi (t) of each sensor signal, a multi-dimensional dynamic threshold θi (t) is set, and the detection conditions are expressed as follows:
Wherein γ and δ represent adjustment parameters, ρij (t) represents a correlation coefficient between signals i and j, Xj (t) represents a sensor signal, μi (t) represents a local mean, σi (t) represents a standard deviation, and θi (t) represents a multidimensional dynamic threshold, respectively;
If the condition |xi(t)-μi(t)|>θi (t) is satisfied, the abnormal point is corrected by considering that the point is abnormal, and the corrected valueExpressed as:
Wherein St represents a normal set of data points within the time window, the weight wk is determined by the similarity between the weight wk and other signals, and k represents the kth sensor;
In the dynamic feature weighting extraction method, the design weight omegai (t) is an adjustment factor based on the activity intensity of poultry in the time period, and is expressed as follows:
Where η represents a conditioning parameter for controlling the influence of heart rate fluctuations on the activity weight.
In one embodiment, the health index Hi is represented as follows:
Where wHR represents the weight of heart rate, wT represents the weight of body temperature, wA represents the weight of activity variance,Represents the average value of the heart rate,The average value of the body temperature is represented,Representing the activity weighted variance.
In one embodiment, the feedback weight adaptive adjustment algorithm dynamically adjusts the weight based on the magnitude of the recent health status deviation Δhi (t), and the specific adjustment formula is:
Where α represents the learning rate, wHE (t) represents the weight of the heart rate at the previous time, wT (t) represents the weight of the body temperature at the previous time, wA (t) represents the weight of the activity variance at the previous time, Δhi (t) represents the recent health state deviation, wHR (t+1) represents the weight of the heart rate after adjustment, wT (t+1) represents the weight of the body temperature after adjustment, and wA (t+1) represents the weight of the activity variance after adjustment.
In one embodiment, the adaptive clustering algorithm based on the behavior feature vector is represented as follows:
the behavior feature vector set of the individual poultry isWhere N is the number of individuals in the population, the particular clustering process is accomplished by minimizing the following clustering objective function:
Wherein K represents the number of clusters, Ck represents the behavior pattern of the kth class, muk represents the center of the class, the objective function J represents the sum of distances from all individual behavior feature vectors to the centers of the classes to which the individual behavior feature vectors belong, and Bj represents the behavior feature vector at a certain moment;
The time series model is modeled by a Markov behavior transformation model.
In one embodiment, the dynamic adjustment of the environmental control strategy is based on a calculated integrated environmental control index D (t), wherein the integrated environmental control index D (t) is represented as follows:
Wherein W represents a weight vector, the weights of the corresponding features, and F (t) represents a multi-modal feature vector;
Defining different threshold intervals according to the comprehensive environment control index, and dynamically adjusting an adjusting strategy of the environment parameters;
comparing the adjustment strategy according to the adjusted environmental parameters with the adjustment strategy before adjustment, and automatically adjusting the weight vector W if the difference exceeds a preset tolerance range, wherein the weight vector W is expressed as follows:
Wherein, alpha represents learning rate, controls the adjustment amplitude of the weight, W (t+1) represents updated weight, W (t) represents current weight, D (t+1) represents updated comprehensive environmental control index, and D (t) represents the previous moment comprehensive environmental control index.
In one embodiment, the composite growth indicator Gi (t) is calculated as follows:
Wherein, alphaW represents the weight coefficient of body weight, alphaS represents the weight coefficient of body shape, alphaH represents the weight coefficient of heart rate, Wi (t) represents the body weight data, Si (t) represents the body shape size data,Representing heart rate data.
In one embodiment, the growth stage differentiation weight distribution algorithm is formulated as follows:
wherein,Representing the values of the environmental parameters at the g-th growth stage,Representing the specific parameter weight under the growth stage, Ec (t) representing the current environmental state and E(g) (t) representing the dynamic environmental demand vector;
the conflict between dynamic environment demands is processed by introducing a multi-objective optimization function based on regularization term, wherein the multi-objective optimization function is expressed as follows:
wherein,Representing a loss function of the environmental regulation,Representing regularization terms between different environmental parameters, λ being a regularization coefficient for balancing the relationship between the environmental parameters;
The feedback mechanism of the growth stage judges whether the current environment is adjusted to meet the growth target or not by monitoring the biological signals after the current environment is adjusted, and the feedback mechanism is expressed as follows:
Wherein γ represents a feedback adjustment coefficient, Δgi (t+1) represents a change in the adjusted growth index, and Δhi (t+1) represents a change in the health state;
The objective function of the comprehensive optimization mechanism of the multi-stage environment control is expressed as follows:
wherein,Representing the parameters of the environment after the optimization,AndRepresenting the actual and expected growth indicators respectively,The standard deviation of the growth at this stage is indicated,Representing regularization terms between the parameters.
In one embodiment, the adaptive multi-objective optimization model is constructed as follows:
Integrating the regulation targets of the system into an adaptive multi-target optimization model, setting a loss function Ln (Θ, t) of each regulation target, wherein Θ is a parameter vector of a control strategy, n represents different regulation targets, and the overall objective function Ltotal (Θ, t) of multi-target optimization is expressed as follows:
Where λn (t) represents the dynamic weight of each target, γ·Θ0∥2 represents a canonical term, Θ0 represents the initial control parameter;
based on the biological signals of the individuals, each individual is assigned a personalized control parameter Θi (t) expressed as follows:
Wherein alpha represents a learning rate, Ji (t) represents a jacobian matrix of the individual i for describing a relationship between the physiological state of the individual and the control parameter,Representing the gradient of the individual loss function, Θi (t+1) represents the updated control parameter.
In a second aspect of the invention, there is provided an adaptive intelligent control system for an avian farming environment, the system comprising:
The data acquisition module is used for installing a miniature sensor on an individual poultry body, acquiring biological signals of the individual poultry body, preprocessing the acquired data, and then designing a multidimensional dynamic boundary abnormality detection algorithm to correct abnormal points to obtain dataData are subjected to design dynamic feature weighting extraction methodExtracting weighted characteristics including heart rate average valueAverage value of body temperatureAnd activity weighted variance
The data analysis modeling module is used for constructing a health state model based on health indexes by using the extracted characteristics, judging the past state of the poultry by calculating the health indexes Hi (t) to obtain the deviation degree delta Hi (t) of the estimated health state, then designing a feedback weight self-adaptive adjustment algorithm to carry out self-adaptive adjustment on the weight in the health state model, grading the health state of the poultry based on the health indexes Hi (t) and the deviation delta Hi (t), and adopting a corresponding environment regulation strategy;
Wherein, the past state of the poultry is judged by calculating the health index Hi (t) to obtain the deviation degree Δhi (t) of the evaluation health state:
If DeltaHi (t) exceeds a preset dynamic threshold value thetaH, indicating that the poultry is in an abnormal or stressed state in the past
The group behavior monitoring and behavior pattern modeling module is used for collecting position coordinate data and speed vector data of birds, mapping the position coordinate data and the speed vector data to behavior feature vectors Bj (t) in a feature space, identifying the behavior patterns of the bird groups according to the behavior feature vectors Bj (t) by adopting a self-adaptive clustering algorithm based on the behavior feature vectors, serializing according to the behavior patterns and constructing a time sequence model based on the behavior patterns to monitor the changes of the group behavior patterns in real time;
The multi-mode data fusion and decision module is used for designing a multi-mode data fusion and weighted decision model to fuse the health index Hi (t), the behavior feature vector Bj (t) and the environmental sensor data to obtain a multi-mode feature vector F (t), generating a comprehensive environmental control index D (t) through a self-adaptive weighting function, dynamically adjusting the environmental control strategy according to the comprehensive environmental control index D (t), and carrying out real-time feedback monitoring after adjustment;
the environment differential regulation and control module is used for constructing a composite growth index Gi (t) for identifying a growth stage according to biological signals, designing a dynamic environment demand vector E(g) (t) according to the composite growth index Gi (t), then designing a growth stage differential weight distribution algorithm to dynamically adjust the priority weight of the dynamic environment demand vector E(g) (t), verifying through a specific growth stage feedback mechanism after each regulation and control, and finally designing a comprehensive optimization mechanism of multi-stage environment control to integrate data of different growth stages;
The system optimization module is used for designing a self-adaptive multi-objective optimization model to dynamically adjust the control strategy according to real-time data, then designing an individual difference self-adaptive control model to distribute individualized control parameters thetai (t) to each individual according to biological signals, dynamically adjusting the control parameters by combining historical data, and designing a multi-level feedback loop and an emergency response mechanism to ensure that the system responds in time under emergency conditions
The beneficial technical effects of the invention are at least as follows:
(1) The invention monitors key physiological indexes such as heart rate, body temperature, activity and the like of birds in real time by installing the biological signal sensor on the individual birds or the activity area thereof. The system analyzes the biological signals by using a deep learning algorithm, and accurately judges the health state and comfort level of the poultry, thereby dynamically adjusting the environmental parameters. Compared with the traditional method, the innovation point obviously improves the precision of environmental regulation, can better respond to the actual demands of poultry individuals, and avoids stress response and health problems caused by improper environmental regulation.
(2) The invention introduces a group behavior monitoring and modeling module. Through computer vision technology, the system can capture and analyze the group behaviors (such as movement, aggregation, dispersion and the like) of the birds in real time, and identify group behavior patterns and abnormal conditions. The system synergistically regulates environmental parameters according to analysis results, such as increasing ventilation or regulating temperature in dense poultry areas, to prevent local overheating or disease transmission. The innovation point obviously improves the response capability of the system to group demands, and ensures the uniformity and stability of the environment.
(3) The invention designs a growth stage identification and differentiation management module. The system recognizes the growth stage of the poultry by monitoring the weight, the body shape and other data of the poultry, and adjusts the illumination intensity, the illumination period and the sound environment in the environment in a targeted manner. For example, during the young bird phase, the system provides a softer light and a quieter environment, while during the egg laying period, the light cycle is optimized and the sound environment is adjusted to promote egg laying efficiency. The innovation point is that the growth quality of poultry is improved, and the cultivation cost is reduced by optimizing energy consumption.
Detailed Description
Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the drawings are illustrative only and are not to be construed as limiting the invention.
In one or more embodiments, as shown in fig. 1, a method for adaptively and intelligently controlling an avian farming environment is disclosed, the method comprising the following steps S1-S6:
s1, installing a miniature sensor on an individual poultry body, collecting biological signals of the individual poultry body, preprocessing the collected data, and then designing a multidimensional dynamic boundary abnormality detection algorithm to correct abnormal points to obtain dataData are subjected to design dynamic feature weighting extraction methodExtracting weighted characteristics including heart rate average valueAverage value of body temperatureAnd activity weighted variance
In particular, to maximize the accuracy and coverage of signal acquisition, the present invention mounts micro sensors at multiple key locations (e.g., chest, wing root and leg) of an individual avian, which sensors are used to acquire biological signals of different dimensions. It is assumed that a heart rate sensor (HR), a temperature sensor (T), and a movement sensor (a) are installed in the chest, the wing root, and the leg, respectively. In addition, to increase reliability of acquisition, redundant sensors (dual sensor redundancy design) are also used at the same location to prevent single sensor failure or error effects. The acquired signal data are recorded as:
HRi(t),Ti(t),Ai(t)(1)
Where i represents the sensor position and t is time.
Aiming at data drift and signal delay possibly generated by different sensors, the invention designs a signal phase alignment algorithm. The algorithm uses a Cross-correlation function (Cross-Correlation Function, CCF) to time-domain align the plurality of sensor signals, the formula:
wherein τij is the optimal time offset between sensors i and j, and after adjustment, the signals are synchronized to the same time reference.
In order to ensure smoothness and continuity of acquired data and particularly solve the problem of large noise in an poultry breeding environment, the invention provides an adaptive weighted Gaussian filter based on a traditional Gaussian filter, and noise levels in different time periods are processed by dynamically adjusting weights. The filtered data xi' (t) is expressed as:
Wherein, the weight wj (t) is adaptively adjusted according to the local noise level and the poultry activity intensity at each moment, and is calculated by adopting the following formula:
here, β is the adjustment parameter, aavg (t) is the average activity for the current time period, and μ and σ are the mean and standard deviation of the gaussian distribution. The term |ai(t)-Aavg (t) | added to this formula is to dynamically adjust the response of the filter according to the severity of the bird's activity, preventing excessive smoothing of the transient noise due to the severe activity.
During the acquisition of the biological signals of the birds, abnormal data may occur due to the movement of the sensor or external interference. In order to accurately identify and correct the abnormal points, the invention provides a multidimensional dynamic boundary abnormality detection algorithm. The algorithm detects outliers by analyzing the correlation between biological signals (heart rate, temperature, activity) in multiple dimensions, determining a dynamic boundary:
First, the local mean μi (t) and standard deviation σi (t) of each sensor signal are calculated, and then a multidimensional dynamic threshold θi (t) is set, where the detection conditions are:
Where γ and δ are adjustment parameters, ρij (t) is the correlation coefficient between signals i and j. If the condition |xi(t)-μi(t)|>θi (t) is satisfied, the point is considered abnormal.
After detecting the abnormality, the multidimensional local weighted regression correction algorithm is applied to correct the abnormal point, and the corrected valueExpressed as:
Where St is the normal set of data points within the time window, the weight wk is determined by its similarity with other signals, similar to the idea of local weighted regression, but optimized in connection with multidimensional data.
After data preprocessing, feature extraction of the data is required to provide efficient input for subsequent steps. The invention provides a dynamic feature weighting extraction method, which not only extracts the traditional mean value and variance, but also weights the signals according to the intensity and stability of the signals. The signal features within each segment L are extracted as follows:
heart rate mean:
Average value of temperature:
activity weighted variance:
Wherein the weight omegai (t) is an adjustment factor based on the activity intensity of the poultry in the time period, and the formula is as follows:
wherein η is a conditioning parameter for controlling the influence of heart rate fluctuations on the activity weight. The processing mode allows physiological responses of birds to be fully considered in the characteristic extraction process, so that the extracted characteristics have more practical significance.
S2, constructing a health state model based on health indexes by using the extracted features, judging the past state of the poultry by calculating the health indexes Hi (t) to obtain the deviation degree delta Hi (t) of the estimated health state, then designing a feedback weight self-adaptive adjustment algorithm to carry out self-adaptive adjustment on the weight in the health state model, classifying the health state of the poultry based on the health indexes Hi (t) and the deviation delta Hi (t), and adopting a corresponding environment regulation strategy.
Wherein, the past state of the poultry is judged by calculating the health index Hi (t) to obtain the deviation degree Δhi (t) of the evaluation health state:
if ΔHi (t) exceeds a preset dynamic threshold ΘH, then it is an indication that the bird was in an abnormal or stressed state in the past.
Specifically, the health state model based on the health index uses the core biosignal features extracted in the first step (such as heart rate averageAverage value of body temperatureAnd activity weighted variance) And establishing a health state model of the poultry. Considering that the influence of the characteristics on the health state of the poultry is different, a health index Hi is constructed by adopting a multi-characteristic fusion mode through weighted summation. The formula for health index Hi is:
Where wHR、wT and wA are the weights of heart rate, body temperature, and activity variance, respectively. These weights are determined from regression analysis of the historical data to ensure optimal predictive power under certain circumstances. By the formula, the biological signal characteristics with different dimensions are integrated into a unified health state index, and a direct basis is provided for subsequent environment regulation and control.
Further, within each time step (e.g., every minute), a health index Hi (t) is calculated over that time period. Then, by averaging the health status with the historyAnd comparing, and evaluating the deviation degree of the current health state. The deviation Δhi (t) is defined as:
wherein,Is an average health index calculated from historical data over a period of time (e.g., the past hour). If ΔHi (t) exceeds a preset dynamic threshold ΘH (set according to the specific farming environment and historical data), then it is indicated that the bird may be in an abnormal or stress state. At this time, the system will start the early warning mechanism and prepare for the next environmental regulation decision.
Further, to cope with changes in the physiological state of birds and dynamic adjustments of the environment, adaptive adjustments to weights wHR、wT and wA in the health state model are required. The invention provides a feedback weight self-adaptive adjustment algorithm which dynamically adjusts weights based on the magnitude of a recent health state deviation delta Hi (t). The specific adjustment formula is as follows:
Where α is the learning rate, and controls the magnitude of the weight adjustment. Through the dynamic adjustment, the system can optimize the contribution of each characteristic to the health state in real time, so that the health index is more fit with the current poultry physiological state.
Finally, the health status of the birds is classified into several classes (e.g., normal, slightly abnormal, severely abnormal) based on the calculated health index Hi (t) and the deviation Δhi (t). According to different levels, the system adopts corresponding environment regulation strategies, such as temperature adjustment, ventilation increase or illumination condition change. For classification purposes, the present invention defines a plurality of thresholds Θ1,Θ2, mapping ΔHi (t) onto corresponding state levels.
Each state level corresponds to a specific environment regulation and control measure, so that the birds can obtain a proper cultivation environment under different health states.
S3, collecting position coordinate data and speed vector data of the poultry, mapping the position coordinate data and the speed vector data to a behavior feature vector Bj (t) in a feature space, identifying a behavior pattern of a poultry group according to the behavior feature vector Bj (t) by adopting an adaptive clustering algorithm based on the behavior feature vector, serializing according to the behavior pattern, and constructing a time sequence model based on the behavior pattern to monitor the change of the behavior pattern of the group in real time.
Further, in order to accurately capture the group behavior dynamics of the birds, a plurality of cameras and motion sensors are arranged in the farm and used for collecting behavior data such as the movement track, the aggregation mode and the activity intensity of the birds in real time. After preprocessing, the data are mapped to a unified behavior feature space. Assume that the collected behavior data includes position coordinates (xj,yj) and a velocity vector 9vx,j,vu,j, where j represents different individuals.
For further analysis, the present invention maps the position data and velocity data to a behavior feature vector Bj (t) in the feature space, respectively, which is expressed as:
The behavior feature vector Bj (t) is used as a basic input for group behavior analysis for subsequent behavior pattern recognition and modeling.
Further, in order to identify the behavior pattern of the bird population, an adaptive clustering algorithm based on behavior feature vectors is employed. Assuming that the behavior feature vector set of the individual poultry isWhere N is the number of individuals in the population. The goal of the clustering algorithm is to divide these behavior feature vectors into classes, each class corresponding to a typical behavior pattern, such as aggregate, scatter, move quickly, etc.
The specific clustering process is implemented by minimizing the following clustering objective function:
Where K is the number of clusters, Ck represents the behavior pattern of the kth class, and μk is the center of the class. The objective function J represents the sum of the distances of all individual behavioral feature vectors to the respective class centers. By optimizing the function, different behavior patterns can be effectively identified. This clustering process dynamically adjusts the value of K to accommodate changes in population behavior over different time periods.
Further, in order to further capture the dynamic change and time dependence of group behaviors, the invention sequences the identified behavior patterns and constructs a time sequence model of the behavior patterns. Assuming a group behavior pattern at time t of Ck (t), the present invention can represent it as a discrete time sequence { Ck (t) }. The time series is modeled by a "markov behavior transformation model".
The state transition probability matrix P of the markov behavior transition model is expressed as:
P=[pij] (19)
where pij＝P(Ci(t+1)|Cj (t)) represents the probability of transitioning from behavior pattern Cj to behavior pattern Ci. Based on the probability matrix, the model can predict future group behavior patterns and identify possible abnormal behavior pattern changes.
Further, based on the results of the time series modeling, the system monitors changes in the group behavior patterns in real time. When a transition probability anomaly of the behavior pattern is detected (i.e., the pij value is significantly below the normal level), the system will automatically trigger the anomaly detection response mechanism. For example, if sudden high density aggregation or irregular movement is detected, the system will record the abnormal event and trigger an environmental regulatory mechanism to adjust temperature, ventilation or lighting conditions.
In addition, the system may also preset an anomaly detection threshold ΘB based on historical behavior pattern data, and immediately take emergency action when the transition probability pij exceeds the threshold.
S4, designing a multi-mode data fusion and weighted decision model to fuse the health index Hi (t), the behavior feature vector Bj (t) and the environmental sensor data to obtain a multi-mode feature vector F (t), generating a comprehensive environmental control index D (t) through a self-adaptive weighting function, dynamically adjusting the environmental control strategy according to the comprehensive environmental control index D (t), and carrying out real-time feedback monitoring after adjustment.
Specifically, based on the health status index Hi (T), the behavior feature Bj (T), and the environmental sensor data (such as the temperature Te (T), the humidity He (T), the illumination intensity Le (T), etc.) generated in the previous steps, the present invention needs to fuse the multidimensional data to generate an optimal control decision suitable for the current cultivation environment. To achieve this, a multimodal data fusion and weighted decision model is proposed.
Specifically, firstly, defining a multi-modal feature vector F (t), and integrating all relevant data sources:
These features are then weighted and summed by an adaptive weighting function to generate the integrated environmental control index D (t). The index is used for guiding specific parameter setting of environment regulation and control, and the calculation formula is as follows:
D(t)=W ·F(t) (21)
Wherein W is a weight vector, and the elements thereof correspond to the weights of the features. The initial value of the weight can be determined by a linear regression method of historical data, and then is adaptively adjusted according to feedback in the running process.
Further, according to the calculated comprehensive environmental control index D (t), the present invention defines different threshold intervals to determine an adjustment strategy of the environmental parameters. For example, the following threshold Θ1、Θ2 may be set to divide different regulation levels:
According to different control grades, the system adjusts environmental parameters such as temperature, humidity, ventilation, illumination and the like. The specific regulation mode can be realized by a linear or nonlinear function, for example, the temperature control function can be designed as follows:
Tc(t+1)＝Tc(t)+βT·(D(t)-Θ1) (23)
Where Tc (T) is the current ambient temperature and βT is the temperature adjustment coefficient. Similarly, humidity, ventilation, illumination, etc. control parameters may also be adjusted according to the integrated environmental control index D (t) and the corresponding adjustment coefficients.
Further, in order to ensure the effectiveness of the environmental regulation strategy, the system performs real-time feedback monitoring after each environmental parameter adjustment. The feedback data includes the new health status indicator Hi (t+1) and the updated behavioral characteristics Bj (t+1). Based on these feedback data, the system recalculates the integrated environmental control index D (t+1) and compares it to D (t) at the previous time. If the variance is outside of a predetermined tolerance range, the system will automatically adjust the weight vector W to optimize future decisions. The specific weight adjustment mechanism is as follows:
wherein alpha is the learning rate, and the amplitude of weight adjustment is controlled. The feedback loop can ensure the self-adaptive adjustment of the system to cope with the dynamic change of the environment and the poultry state, and realize the long-term environment control optimization.
Further, finally, in order to improve the robustness and adaptability of the system, the invention designs a multi-level environment control mechanism. When a system is detected to fail to continuously adjust for a plurality of times (i.e. the feedback data indicates that D (t+1) is not changing towards the expected direction), the system automatically enters an emergency state and a standby control strategy is started. Such strategies are typically based on successful cases in historical data, by directly invoking past environmental parameter settings to maintain system stability until the problem is resolved or external intervention. The control function in the emergency state can be expressed as:
Fbackup(t)＝Fhistory(tbest) (25)
Wherein Fhistory(tbest) is the feature vector corresponding to the best control effect in the history data. By the mechanism, the system can still maintain the stability of the environment when facing complex or unknown problems, and ensure the health and the cultivation efficiency of poultry.
S5, constructing a composite growth index Gi (t) for identifying a growth stage according to biological signals, designing a dynamic environment demand vector E(g) (t) according to the composite growth index Gi (t), then designing a growth stage differentiation weight distribution algorithm to dynamically adjust the priority weight of the dynamic environment demand vector E(g) (t), verifying through a specific growth stage feedback mechanism after each regulation, and finally designing a comprehensive optimization mechanism of multi-stage environment control to integrate data of different growth stages.
Specifically, in order to achieve differentiated environmental control based on the growth phase of birds, it is first necessary to accurately identify the current growth phase of birds. The biological signal features (such as body weight Wi (t), body size Si (t), heart rate) have been extracted in step 1Etc.), the present invention combines these features to construct a composite growth indicator Gi (t) that identifies the growth phase. Specifically, the calculation formula of Gi (t) is:
Wherein, alphaw、αS and alphaH are weight coefficients of body weight, body type and heart rate respectively, and are obtained through regression analysis of historical data or experimental data calibration. After standardized treatment, the composite growth index Gi (t) may be used to divide different growth phases (e.g., young bird phase, breeding phase, egg producing phase, etc.), each phase employing a specific environmental control strategy.
To closely relate the identified growth phase to the environmental demand, a dynamic environmental demand vector E(g) (T) is defined that contains the temperature demand Tg (T), the humidity demand Hg (T), the lighting demand Lg (T), and the ventilation demand Vg (T), which are modeled based on the specific demands of the current phase. For example:
Wherein f (Gi (t), G) is a nonlinear mapping function that converts the composite growth index and growth phase into specific environmental requirements.
Further, after identifying the growth phase, the system compares the dynamic environmental demand with the current environmental state (represented by the environmental control index D (t) in step 4) and determines the environmental parameters that need to be adjusted. In order to realize accurate regulation, a 'differential weight distribution algorithm in growth stage' is provided. The algorithm dynamically adjusts the priority weight of each environmental parameter according to the requirements of the current growth stage so as to ensure that the environmental regulation and control can meet the requirements of the specific growth stage.
The adjustment formula of the environmental control parameter is:
wherein,Is the environmental parameter value at the g-th growth stage,Is a specific parameter weight at this growth stage, and Ec (t) is the current environmental state. Through the formula, the system adjusts the current environmental state to a state that best meets the growth stage requirements. The adjustment process considers the priority of each parameter, thereby realizing the fine management of environment regulation.
Furthermore, to address possible conflicts between dynamic environmental demands (e.g., high temperature and high humidity may have negative effects), a multi-objective optimization function based on regularization terms was introduced:
wherein,Is a loss function of the environmental regulation and control,Is a regularization term between different environmental parameters, λ is a regularization coefficient used to balance the relationship between environmental parameters. By minimizing the loss function, the system can find an optimal environmental regulation strategy under the condition of multi-parameter conflict.
Further, to ensure the effectiveness and adaptability of environmental regulation, the system is verified by a specific growth stage feedback mechanism after each regulation. The mechanism is implemented by monitoring the biological signal (such as new heart rate) after the current environment is adjustedBody temperature) And a growth index Gi (t+1) to determine whether the current environment is adjusted to meet the growth target.
The specific feedback adjustment formula is:
Where γ is a feedback adjustment coefficient, Δgi (t+1) represents a change in the growth index after adjustment, and Δhi (t+1) represents a change in the health state. Through the feedback mechanism, the system can dynamically adjust the priority weight of the environmental parameters, and ensures that the growth index develops towards the expected direction.
Finally, in order to ensure the consistency and optimization of the environmental control in the whole growth period, a comprehensive optimization mechanism of the multi-stage environmental control is designed. The mechanism optimizes the control strategy of the current growth stage by integrating the data of different growth stages and utilizing the historical data. By comparing the growth curves at different stages with the environmental control parameters, the system can identify the optimal control path and apply it in the current environment.
The objective function of the comprehensive optimization is as follows:
wherein,Is the environmental parameter after the optimization,AndThe actual and expected growth indicators are respectively,Is the standard deviation of the growth at this stage,Is a regularization term between parameters. By minimizing the objective function, the system can achieve comprehensive optimization of different growth stages, ensuring continuity and consistency of environmental control.
S6, designing a self-adaptive multi-objective optimization model to dynamically adjust a control strategy according to real-time data, then designing an individual difference self-adaptive control model to distribute individualized control parameters thetai (t) for each individual according to biological signals, dynamically adjusting the control parameters by combining historical data, and designing a multi-level feedback loop and emergency response mechanism to ensure timely response of the system under emergency conditions.
Specifically, in the previous steps, the system gradually establishes a complex and flexible intelligent control system by collecting multidimensional data in real time, constructing a health state model, analyzing group behaviors, controlling environment and regulating and controlling the growth stage. However, as birds grow and the external environment continues to change, the system needs to further increase the adaptive capacity, dynamically adjust the control strategy to optimize the long-term farming effect. In a sixth step, an adaptive multi-objective optimization model is proposed to achieve this objective.
First, the present invention integrates multiple regulatory targets (e.g., temperature, humidity, illumination intensity, etc.) of the system into one adaptive multi-target optimization model. The loss function Lk (Θ, t) of each regulation target is set, wherein Θ is a parameter vector of a control strategy, and k represents different regulation targets, such as temperature control, humidity regulation and the like. The overall objective function Ltotal (Θ, t) for multi-objective optimization can be expressed as a weighted sum of the sub-objectives:
Where λk (t) is the dynamic weight of each target, γ·the/- Θ0∥2 is a regularized term used to limit the range of variation of the parameters, avoid excessive excitation of system regulation, and Θ0 is the initial control parameter. The optimization model unifies the regulation and control targets of the system under an optimization framework by integrating multidimensional information (such as health state indexes Hi (t), behavior characteristics Bj (t) and environment control indexes D (t)) obtained in the previous steps, so that balance and coordination among the targets are realized. Through self-adaptive adjustment of the weight lambdak (t), the system can dynamically adapt to the environment and the change of the poultry demand, and long-term stable and efficient control is realized.
Further, in the cultivation process, physiological states and environmental requirements of different individuals are different, and the system needs to have personalized regulation and control capability. The invention designs an individual difference self-adaptive control model based on the individual characteristics (such as heart rate)Body temperatureActivity level) Each individual is assigned a personalized control parameter Θi (t).
In particular, the present invention uses a nonlinear feedback mechanism to adjust individual control parameters. Assuming that the current control parameter of the individual i is Θi (t), the formula of feedback regulation is:
where alpha is the learning rate, Ji (t) is the jacobian matrix of the individual i describing the relationship between the individual's physiological state and the control parameters,Is the gradient of the individual loss function. This nonlinear feedback mechanism allows the system to adjust its control parameters based on real-time feedback from the individual, enabling fine management. By means of the model, the system can identify and adapt to differences among individuals, and particularly can conduct personalized regulation and control on individuals with certain health states or special growth phases. The mechanism is tightly combined with biological signal monitoring and health status modeling in the previous step, so that the flexibility and response capability of the system are further improved.
Further, during long-term farming, the system needs to constantly learn from historical data to promote future control strategies. For this reason, a "history-driven long-acting learning mechanism" is introduced that optimizes the control strategy of the system by analyzing and summarizing the historical farming data. The invention constructs a historical data loss function Lhistory (Θ) expressed as:
wherein,Representing the desired operation, T is the historical time period, R (Θ) is a regularization term, and δ is its weight. By minimizing this historical loss function, the system is able to optimize the parameter Θ, ensuring that an optimal control strategy is maintained during long-term operation. The mechanism combines the multidimensional data and the control experience accumulated in the previous steps, and improves the effectiveness and the robustness of a future control strategy by summarizing and optimizing historical data. The learning mechanism based on the historical data can capture the long-term trend of the environment and the physiological state, and provides a solid foundation for the long-term stable operation of the system.
Further, in order to ensure that the system can respond in time under the emergency, a multi-level feedback loop and emergency response mechanism is designed. The mechanism automatically enters a feedback loop after each environmental regulation based on the real-time data generated in the previous step (e.g., change in health state Δhi (t), abrupt change in environmental state Ep (t), etc.), evaluates the current regulation effect, and decides whether to trigger an emergency response.
If multiple regulatory failures or significant deterioration of environmental conditions (e.g., a sharp rise in temperature) are detected, the system will enter an emergency state, enabling a backup strategy. The decision model of the emergency response is expressed as:
Θemergency＝Θhistory+ξ·ΔE (35)
Where Θemergency is a control parameter in an emergency state, Θhistory is a history optimization strategy, ζ is an adjustment coefficient, and ΔE is an emergency adjustment amount of an environmental parameter. The formula ensures that the system can react quickly when facing emergency, and the stability of the environment is maintained. The multi-level feedback loop and the emergency response mechanism not only can rapidly cope with environmental changes, but also can be organically combined with the previous individual difference regulation and control, self-adaptive learning and historical data driving mechanisms to form a complete closed-loop control system. The system plays a key role in protecting in the whole patent proposal, and ensures that the system can keep high-efficiency and stable operation under various conditions.
In one or more embodiments, as shown in fig. 2, an adaptive intelligent control system for an avian farming environment is disclosed, the system comprising:
The data acquisition module 101 is used for installing a micro sensor on an individual poultry, acquiring biological signals of the individual poultry, preprocessing the acquired data, and then designing a multidimensional dynamic boundary abnormality detection algorithm to correct abnormal points to obtain dataData are subjected to design dynamic feature weighting extraction methodExtracting weighted characteristics including heart rate average valueAverage value of body temperatureAnd activity weighted variance
The data analysis modeling module 102 is configured to construct a health state model based on a health index by using the extracted features, determine a past state of the bird by calculating a health index Hi (t) to obtain a deviation degree Δhi (t) of an estimated health state, then design a feedback weight adaptive adjustment algorithm to adaptively adjust weights in the health state model, then rank the health state of the bird based on the health index Hi (t) and the deviation Δhi (t), and adopt a corresponding environment regulation strategy;
Wherein, the past state of the poultry is judged by calculating the health index Hi (t) to obtain the deviation degree Δhi (t) of the evaluation health state:
If DeltaHi (t) exceeds a preset dynamic threshold value thetaH, indicating that the poultry is in an abnormal or stressed state in the past
The group behavior monitoring and behavior pattern modeling module 103 is used for collecting position coordinate data and speed vector data of birds, mapping the position coordinate data and the speed vector data to behavior feature vectors Bj (t) in a feature space, identifying the behavior patterns of the bird groups according to the behavior feature vectors Bj (t) by adopting an adaptive clustering algorithm based on the behavior feature vectors, serializing according to the behavior patterns and constructing a time sequence model based on the behavior patterns to monitor the changes of the group behavior patterns in real time;
The multi-mode data fusion and decision module 104 is configured to design a multi-mode data fusion and weighted decision model to fuse the health index Hi (t), the behavior feature vector Bj (t) and the environmental sensor data to obtain a multi-mode feature vector F (t), then generate a comprehensive environmental control index D (t) through a self-adaptive weighting function, dynamically adjust the environmental control strategy according to the comprehensive environmental control index D (t), and perform real-time feedback monitoring after adjustment;
the environment differential regulation and control module 105 is used for constructing a composite growth index Gi (t) for identifying a growth stage according to biological signals, designing a dynamic environment demand vector E(g) (t) according to the composite growth index Gi (t), then designing a growth stage differential weight distribution algorithm to dynamically adjust the priority weight of the dynamic environment demand vector E(g) (t), verifying through a specific growth stage feedback mechanism after each regulation and control, and finally designing a comprehensive optimization mechanism of multi-stage environment control to integrate data of different growth stages;
The system optimization module 106 is configured to design an adaptive multi-objective optimization model to dynamically adjust the control strategy according to the real-time data, then design an individual difference adaptive control model to allocate individual control parameters Θi (t) to each individual according to the biological signal, dynamically adjust the control parameters in combination with the historical data, and design a multi-level feedback loop and an emergency response mechanism to ensure timely response of the system in an emergency.
It should be noted that, the specific workflow of the adaptive intelligent control system for poultry farming environment provided by the embodiment of the present invention is the same as the workflow of the adaptive intelligent control method for poultry farming environment described in the above embodiment, and will not be described herein.
Compared with the prior art, the poultry breeding environment self-adaptive intelligent control system provided by the embodiment of the invention is characterized in that a miniature sensor is arranged on an individual poultry body, biological signals of the individual poultry body are collected, the collected data are preprocessed, and then a multidimensional dynamic boundary abnormality detection algorithm is designed to correct abnormal points to obtain the dataData are subjected to design dynamic feature weighting extraction methodExtracting weighted characteristics including heart rate average valueAverage value of body temperatureAnd activity weighted varianceConstructing a health state model based on health indexes by using the extracted characteristics, judging the past state of poultry by calculating the health index Hi (t) to obtain the deviation degree delta Hi (t) of the evaluated health state, then designing a feedback weight self-adaptive adjustment algorithm to carry out self-adaptive adjustment on the weight in the health state model, grading the health state of the poultry based on the health index Hi (t) and the deviation delta Hi (t), and adopting a corresponding environment regulation strategy, wherein the past state of the poultry is judged by calculating the health index Hi (t) to obtain the deviation degree delta Hi (t) of the evaluated health state, if delta Hi (t) exceeds a preset dynamic threshold value thetaH, the past state of the poultry is indicated, acquiring position coordinate data and speed vector data of the poultry, mapping the position coordinate data and the speed vector data to a behavior characteristic vector Bj (t) in a characteristic space, identifying a behavior pattern of a poultry group according to the behavior characteristic vector Bj (t) by adopting a self-adaptive clustering algorithm based on the behavior characteristic vector, constructing a behavior pattern sequence, and fusing the behavior pattern sequence with the behavior pattern model based on a time sequence to design a decision-weighted mode of the behavior pattern and the model (i) to be changed in real time-weighted mode of the health state model (i), The method comprises the steps of fusing behavior feature vectors Bj (t) with environment sensor data to obtain multi-mode feature vectors F (t), generating comprehensive environment control indexes D (t) through a self-adaptive weighting function, dynamically adjusting environment control strategies according to the comprehensive environment control indexes D (t), carrying out real-time feedback monitoring after adjustment, constructing composite growth indexes Gi (t) for identifying growth stages according to biological signals, designing dynamic environment demand vectors E(g) (t) according to the composite growth indexes Gi (t), dynamically adjusting priority weights of the dynamic environment demand vectors E(g) (t) according to a differential weight distribution algorithm of the growth stages, verifying through a specific growth stage feedback mechanism after each adjustment, finally designing a multi-stage environment control comprehensive optimization mechanism to integrate data of different growth stages, designing a self-adaptive multi-target optimization model, dynamically adjusting the control strategies according to real-time data, designing individual differential self-adaptive control models, distributing personalized control parameters thetai (t) for each individual according to biological signals, dynamically adjusting the control parameters according to the historical data, and designing a multi-level feedback cycle and an emergency response system under emergency response.
The embodiment of the invention also provides an adaptive intelligent control device for the poultry farming environment, which comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein the steps in the adaptive intelligent control method for the poultry farming environment are realized when the processor executes the computer program, such as steps S1-S6 in the embodiment of the adaptive intelligent control method for the poultry farming environment, for example, as shown in fig. 1, or the functions of the modules in the embodiments of the systems are realized when the processor executes the computer program.
The computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to accomplish the present invention, for example. The one or more modules may be a series of computer program instruction segments capable of performing specific functions for describing the execution of the computer program in the poultry farming environment adaptive intelligent control device.
The poultry breeding environment self-adaptive intelligent control device can be a desktop computer, a notebook computer, a palm computer, a cloud server and other computing devices. The poultry farming environment adaptive intelligent control device may include, but is not limited to, a processor, a memory. Those skilled in the art will appreciate that the poultry farming environment adaptive intelligent control device may also include input and output devices, network access devices, buses, and the like.
The processor may be a central processing unit (CentralProcessangUnat, CPU), but may also be other general purpose processors, digital signal processors (DagatalSagnalProcessor, DSP), application specific integrated circuits (ApplacataonSpecafacAntegratedCarcuat, ASAC), field programmable gate arrays (Faeld-ProgrammableGateArray, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or the like. The general processor can be a microprocessor or any conventional processor, etc., and the processor is a control center of the adaptive intelligent control device for poultry farming environment, and various interfaces and lines are used to connect various parts of the adaptive intelligent control device for poultry farming environment.
The memory can be used for storing the computer program and/or the module, and the processor can realize various functions of the poultry farming environment self-adaptive intelligent control device by running or executing the computer program and/or the module stored in the memory and calling the data stored in the memory. The memory may mainly include a storage program area which may store an operating system, an application program required for at least one function, and the like, and a storage data area which may store data created according to the operation of the air conditioner controller, and the like. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, smart memory card (SMARTMEDAACARD, SMC), secure digital (SecureDagatal, SD) card, flash memory card (FLASHCARD), at least one disk storage device, flash memory device, or other volatile solid-state storage device.
The module integrated by the poultry farming environment self-adaptive intelligent control device can be stored in a computer readable storage medium if the module is realized in the form of a software functional unit and sold or used as an independent product. Based on such understanding, the present invention may implement all or part of the flow of the method of the above embodiment, or may be implemented by a computer program to instruct related hardware, where the computer program may be stored in a computer readable storage medium, and when the computer program is executed by a processor, the computer program may implement the steps of each of the method embodiments described above. Wherein the computer program comprises computer program code which may be in source code form, object code form, executable file or some intermediate form etc. The computer readable medium may include any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer memory, a Read-only memory (ROM), a random access memory (RAM, randomAccessMemory), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-only memory (ROM), a random access memory (RandomAccessMemory, RAM), or the like.
While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that changes and modifications may be made without departing from the principles of the invention, such changes and modifications are also intended to be within the scope of the invention.