Cow feeding environment comprehensive control system based on artificial intelligence and control method thereof

Abstract

The invention discloses a cattle raising environment comprehensive control system based on artificial intelligence, and the system comprises a plurality of temperature and humidity sensors which are used for monitoring the temperature and humidity of a cattle house in real time; the plurality of illumination intensity sensors are used for monitoring the illumination intensity of the cowshed in real time; the gas concentration sensors are used for monitoring the concentration of carbon dioxide, ammonia gas and hydrogen sulfide in the cowshed in real time; the air conditioning unit and the humidifier are used for controlling the temperature and the humidity of the cowshed; the electric window shade and the light supplementing lamp are used for controlling the illumination intensity of the cowshed; the fresh air fan is used for carrying out air renewal on the cowshed; and the central controller is in communication connection with the temperature and humidity sensor, the illumination intensity sensor, the gas concentration sensor, the air conditioning unit, the humidifier, the electric window shade, the light supplementing lamp and the fresh air fan and is used for comprehensively controlling the environment of the cowshed according to the collected environmental parameters of the cowshed. The defects in the prior art can be overcome, and the regulation and control effect of the control system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent livestock breeding, and in particular to an artificial intelligence-based cattle breeding environment integrated control system and a control method thereof. Background Art

[0002] With the popularization of the Internet of Things and intelligent technology, the livestock farming industry has gradually reduced the intensity of manual work and optimized the livestock farming environment through automated monitoring and control systems. Since there are many parameters in the livestock farming environment and the amount of data obtained by real-time monitoring is large, a commonly used method in the prior art is to determine and predict the environmental parameters through a particle swarm model to obtain the current optimal adjustment target, and then regulate the breeding environment through a control system. However, since the theoretical optimal adjustment target is not the current optimal adjustment target (because livestock breeding also needs to consider the livestock's own adaptability to the amplitude and speed of changes in environmental parameters and the mutual influence between different parameters), the particle swarm optimization process is prone to slow convergence and optimization, so the existing particle swarm model still needs to be improved and optimized. Summary of the invention

[0003] The purpose of the present invention is to provide an artificial intelligence-based integrated control system for cattle breeding environment and a control method thereof to solve the problems raised in the above-mentioned background technology.

[0004] An artificial intelligence-based integrated control system for cattle breeding environment, comprising:

[0005] Several temperature and humidity sensors are installed in different locations of the cowshed to monitor the temperature and humidity of the cowshed in real time;

[0006] Several light intensity sensors are installed at different locations in the cowshed to monitor the light intensity in the cowshed in real time;

[0007] Several gas concentration sensors are installed at different locations in the cowshed to monitor the concentrations of carbon dioxide, ammonia and hydrogen sulfide in the cowshed in real time;

[0008] Air conditioning units and humidifiers are used to control the temperature and humidity of the cowshed;

[0009] Electric blackout curtains and fill lights to control the light intensity in the cowshed;

[0010] Fresh air fan, used to renew the air in the cowshed;

[0011] The central controller is respectively connected to the temperature and humidity sensors, light intensity sensors, gas concentration sensors, air conditioning units, humidifiers, electric blackout curtains, fill lights and fresh air fans for comprehensive control of the cowshed environment based on the collected cowshed environmental parameters.

[0012] Preferably, the air-conditioning unit and the fresh air fan are installed as an integrated structure, at least two humidifiers are arranged in the cowshed, at least one humidifier is installed at the air outlet of the air-conditioning unit, and at least one humidifier is installed at a distance of more than 5m from the air outlet and air inlet of the air-conditioning unit.

[0013] Preferably, the number of temperature and humidity sensors and gas concentration sensors is equal, and each temperature and humidity sensor is installed at the same position as a gas concentration sensor.

[0014] A control method of the above-mentioned cattle breeding environment integrated control system based on artificial intelligence comprises the following steps:

[0015] A. The temperature and humidity sensor monitors the temperature and humidity of the cowshed in real time, the light intensity sensor monitors the light intensity of the cowshed in real time, and the gas concentration sensor monitors the concentration of carbon dioxide, ammonia and hydrogen sulfide in the cowshed in real time. The central controller collects the above cowshed environmental parameters;

[0016] B. The central controller pre-processes the collected cowshed environmental parameters and inputs them into the random forest-particle swarm model to obtain the target values ​​for adjusting the cowshed environmental parameters;

[0017] C. The central controller adjusts the target value according to the environmental parameters of the cowshed to control the air-conditioning unit, humidifier, electric blackout curtains, fill light and fresh air fan.

[0018] Preferably, preprocessing the collected cattle house environmental parameters comprises the following steps:

[0019] The collected discrete parameters of the same type are fitted, and a parameter curve is obtained for each type of parameter; the fluctuation standard deviation threshold of each parameter curve is set, and the parameter curve is segmented. Each curve segment maximizes its own length while ensuring that its fluctuation standard deviation is less than the fluctuation standard deviation threshold. The average value of each curve segment is calculated, and each average value and its corresponding parameter curve are encapsulated into a data packet.

[0020] Preferably, the particle swarm-random forest model includes a random forest prediction module and a particle swarm prediction module; a data packet is input into the random forest prediction module, each random tree outputs a flight range of all particles, and the flight ranges of all random tree output ends are combined as the output of the random forest prediction module; a data packet is input into the particle swarm prediction module, particles perform random flight updates according to the flight range output by the random forest prediction model, the particle swarm prediction module outputs the optimal cowshed environmental parameter adjustment target value, and then the output of the random tree is weightedly corrected according to the current optimal cowshed environmental parameter adjustment target value.

[0021] Preferably, the cowshed environmental parameter combinations are classified according to the feeding requirements of the cattle herd, and the boundary conditions of each cowshed environmental parameter combination are used as the splitting features of the random tree to generate the random tree in the random forest prediction module.

[0022] As a preference, the particle swarm prediction module is initialized before prediction, and the objective function of the particle is Min(F)=|P0-P|×|P b -P|, where P is the current position of the particle, P0 is the initial position of the particle, and P b It is the particle position corresponding to the optimal combination of theoretical cowshed environmental parameters obtained according to the feeding requirements of the cattle herd; after the particle swarm prediction module outputs the optimal cowshed environmental parameter adjustment target value, the output of the random tree is weighted and corrected, and the weighted value is inversely proportional to the distance between the geometric center of the flight range output by the random tree and the optimal cowshed environmental parameter adjustment target value output by the particle swarm prediction module.

[0023] Compared with the prior art, the beneficial effect of the present invention is that: on the basis of the existing particle swarm algorithm, by optimizing the objective function and using the random forest model to limit the flight range of particles, the particle swarm model can quickly and accurately obtain the current optimal regulation target, thereby improving the regulation effect of the control system. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 The system schematic diagram of a specific implementation mode of the present invention. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] An artificial intelligence-based integrated control system for cattle breeding environment, comprising:

[0027] Several temperature and humidity sensors 1 are installed at different locations in the cowshed to monitor the temperature and humidity of the cowshed in real time;

[0028] Several light intensity sensors 2 are installed at different locations in the cowshed to monitor the light intensity in the cowshed in real time;

[0029] Several gas concentration sensors 3 are installed at different locations in the cowshed to monitor the concentrations of carbon dioxide, ammonia and hydrogen sulfide in the cowshed in real time;

[0030] The air conditioning unit 4 and the humidifier 5 are used to control the temperature and humidity of the cowshed;

[0031] The electric blackout curtain 6 and the fill light 7 are used to control the light intensity of the cowshed;

[0032] Fresh air fan 8, used for renewing the air in the cowshed;

[0033] The central controller 9 is respectively connected to the temperature and humidity sensor 1, the light intensity sensor 2, the gas concentration sensor 3, the air conditioning unit 4, the humidifier 5, the electric blackout curtain 6, the fill light 7 and the fresh air fan 8 for comprehensive control of the cowshed environment according to the collected cowshed environmental parameters.

[0034] This system uses random forest-particle swarm algorithm to predict environmental parameters. It includes the following steps:

[0035] A. The temperature and humidity sensor 1 monitors the temperature and humidity of the cowshed in real time, the light intensity sensor 2 monitors the light intensity of the cowshed in real time, the gas concentration sensor 3 monitors the concentration of carbon dioxide, ammonia and hydrogen sulfide in the cowshed in real time, and the central controller 9 collects the above-mentioned cowshed environmental parameters;

[0036] B. The central controller 9 pre-processes the collected cowshed environmental parameters and inputs them into the random forest-particle swarm model to obtain the target value of the cowshed environmental parameters;

[0037] C. The central controller 9 adjusts the target value according to the environmental parameters of the cowshed to control the air-conditioning unit 4, the humidifier 5, the electric blackout curtain 6, the fill light 7 and the fresh air fan 8 for regulation.

[0038] The existing particle swarm algorithm cannot effectively limit the flight range of particles according to the multi-dimensional data information contained in the particles. Even if it is limited, it is mostly given a limited range based on empirical values, which results in the inability to quickly and directly obtain the best optimization range during the optimization process. The present invention solves this problem by introducing a random forest model. Although there are also applications that combine random forest algorithms with particle swarm algorithms in the prior art, the algorithm logic is to use the particle swarm algorithm to select a feature subset of the random forest model to simplify the generation operation process of the random forest. This is completely different from the algorithm we designed.

[0039] In order to make the data suitable for the calculation of the random forest model and the particle swarm model, the present invention preprocesses the collected cattle house environmental parameters, including the following steps:

[0040] The collected discrete parameters of the same type are fitted, and a parameter curve is obtained for each type of parameter; the fluctuation standard deviation threshold of each parameter curve is set, and the parameter curve is segmented. Each curve segment maximizes its own length while ensuring that its fluctuation standard deviation is less than the fluctuation standard deviation threshold. The average value of each curve segment is calculated, and each average value and its corresponding parameter curve are encapsulated into a data packet.

[0041] The data packet encapsulated by the present invention includes not only the average value (i.e., characteristic parameter) but also the parameter curve related thereto. After the data packet is input into the random forest model, the parameter curve is used to predict the flight range of the particle. After the data packet is input into the particle swarm model, the combination of the average values ​​is used as the initial state of the particle to perform optimization calculation.

[0042] The particle swarm-random forest model includes a random forest prediction module and a particle swarm prediction module.

[0043] The data packet is input into the random forest prediction module, and each random tree outputs a flight range of all particles. The flight ranges of all random tree outputs are combined as the output of the random forest prediction module. The cowshed environmental parameter combinations are classified according to the feeding requirements of the cattle herd, and the boundary conditions of each cowshed environmental parameter combination are used as the splitting features of the random tree to generate the random trees in the random forest prediction module.

[0044] The data packet is input into the particle swarm prediction module, and the particles are randomly updated according to the flight range output by the random forest prediction model. The particle swarm prediction module outputs the optimal cattle shed environmental parameter adjustment target value, and then the output of the random tree is weighted and corrected according to the current optimal cattle shed environmental parameter adjustment target value. The particle swarm prediction module is initialized before prediction, and the objective function of the particle is set to Min(F) = |P0-P|×P b -P|, where P is the current position of the particle, P0 is the initial position of the particle, and P b The particle position corresponding to the theoretical optimal combination of cattle house environmental parameters obtained according to the feeding requirements of the cattle herd. By optimizing the objective function, the particle swarm can not only have the limitation of the theoretical optimal target value, but also the limitation of the parameter adjustment range. At the same time, the optimized search space range is obtained through the random forest prediction model, realizing the fast and accurate convergence of the particle swarm algorithm.

[0045] After the particle swarm prediction module outputs the optimal target value for adjusting the environmental parameters of the cattle shed, the output of the random tree is weighted and corrected, and the weighted value is inversely proportional to the distance between the geometric center of the flight range output by the random tree and the optimal target value for adjusting the environmental parameters of the cattle shed output by the particle swarm prediction module. By weighting the output of the random tree, the entire random forest prediction model can be kept synchronized with the environmental parameters at the time of the next round of prediction calculation, reducing the error caused by the inability of the random forest prediction model to be retrained in real time due to real-time changes in environmental parameters.

Claims

1. An artificial intelligence-based integrated control system for cattle breeding environment, characterized in that: include, A plurality of temperature and humidity sensors (1) are installed at different locations in the cowshed to monitor the temperature and humidity of the cowshed in real time; A plurality of light intensity sensors (2) are installed at different locations of the cowshed to monitor the light intensity of the cowshed in real time; A plurality of gas concentration sensors (3) are installed at different locations in the cowshed to monitor the concentrations of carbon dioxide, ammonia and hydrogen sulfide in the cowshed in real time; The air conditioning unit (4) and the humidifier (5) are used to control the temperature and humidity of the cowshed; The electric blackout curtain (6) and the fill light (7) are used to control the light intensity of the cowshed; A fresh air fan (8) is used to renew the air in the cowshed; The central controller (9) is respectively connected to the temperature and humidity sensor (1), the light intensity sensor (2), the gas concentration sensor (3), the air conditioning unit (4), the humidifier (5), the electric blackout curtain (6), the fill light (7) and the fresh air blower (8) for comprehensive control of the cowshed environment according to the collected cowshed environmental parameters.

2. The artificial intelligence-based cattle breeding environment integrated control system according to claim 1 is characterized in that: The air conditioning unit (4) and the fresh air blower (8) form an integrated structure for installation. At least two humidifiers (5) are arranged in the cowshed, and at least one humidifier (5) is installed at the air outlet of the air conditioning unit (4). The installation position of at least one humidifier (5) is greater than 5m away from the air outlet and the air inlet of the air conditioning unit (4).

3. The artificial intelligence-based cattle breeding environment integrated control system according to claim 1 is characterized in that: The number of the temperature and humidity sensors (1) and the gas concentration sensors (3) is equal, and each temperature and humidity sensor (1) and a gas concentration sensor (3) are installed at the same position.

4. A control method for an artificial intelligence-based cattle breeding environment integrated control system according to any one of claims 1 to 3, characterized in that The steps include: A. The temperature and humidity sensor (1) monitors the temperature and humidity of the cowshed in real time, the light intensity sensor (2) monitors the light intensity of the cowshed in real time, the gas concentration sensor (3) monitors the concentration of carbon dioxide, ammonia and hydrogen sulfide in the cowshed in real time, and the central controller (9) collects the above-mentioned cowshed environmental parameters; B. The central controller (9) pre-processes the collected cowshed environmental parameters and inputs them into the random forest-particle swarm model to obtain the target values ​​for adjusting the cowshed environmental parameters; C. The central controller (9) controls the air conditioning unit (4), the humidifier (5), the electric blackout curtain (6), the fill light (7) and the fresh air fan (8) according to the target value of the environmental parameters of the cowshed.

5. The control method of the cattle breeding environment integrated control system based on artificial intelligence according to claim 4 is characterized in that: The preprocessing of the collected cattle house environmental parameters includes the following steps: The collected discrete parameters of the same type are fitted, and a parameter curve is obtained for each type of parameter; the fluctuation standard deviation threshold of each parameter curve is set, and the parameter curve is segmented. Each curve segment maximizes its own length while ensuring that its fluctuation standard deviation is less than the fluctuation standard deviation threshold. The average value of each curve segment is calculated, and each average value and its corresponding parameter curve are encapsulated into a data packet.

6. The control method of the cattle breeding environment integrated control system based on artificial intelligence according to claim 5 is characterized by: The particle swarm-random forest model includes a random forest prediction module and a particle swarm prediction module; a data packet is input into the random forest prediction module, each random tree outputs a flight range of all particles, and the flight ranges of all random tree output ends are unioned as the output of the random forest prediction module; a data packet is input into the particle swarm prediction module, and particles perform random flight updates according to the flight range output by the random forest prediction model, and the particle swarm prediction module outputs the optimal cowshed environmental parameter adjustment target value, and then the output of the random tree is weightedly corrected according to the current optimal cowshed environmental parameter adjustment target value.

7. The control method of the cattle breeding environment integrated control system based on artificial intelligence according to claim 6 is characterized by: The cowshed environmental parameter combinations are classified according to the feeding requirements of the cattle herd, and the boundary conditions of each cowshed environmental parameter combination are used as the splitting features of the random tree to generate the random trees in the random forest prediction module.

8. The control method of the cattle breeding environment integrated control system based on artificial intelligence according to claim 7 is characterized by: The particle swarm prediction module is initialized before prediction, and the objective function of the particle is Min(F) = |P0-P|×|P b -P|, where P is the current position of the particle, P0 is the initial position of the particle, and P b It is the particle position corresponding to the optimal combination of theoretical cowshed environmental parameters obtained according to the feeding requirements of the cattle herd; after the particle swarm prediction module outputs the optimal cowshed environmental parameter adjustment target value, the output of the random tree is weighted and corrected, and the weighted value is inversely proportional to the distance between the geometric center of the flight range output by the random tree and the optimal cowshed environmental parameter adjustment target value output by the particle swarm prediction module.

Images