Intelligent sheep house temperature and humidity regulation and control system based on artificial intelligence

Abstract

The invention discloses a sheep house temperature and humidity intelligent regulation and control system based on artificial intelligence, and aims at solving the problems that a traditional regulation and control system is generally poor in partition control capability, low in response speed, lack of pertinence and the like. The system is composed of temperature and humidity sensors, a data acquisition unit, a temperature and humidity adjustment execution device and an artificial intelligence control unit which are arranged in different functional zones in a sheep house. Wherein the temperature and humidity sensor acquires environmental parameters and is connected with the data acquisition unit through a bus; the data acquisition unit summarizes multi-point sensing data and then transmits the summarized multi-point sensing data to the artificial intelligence control unit at a fixed frequency. According to the invention, an adaptive partition control scheme driven by an artificial intelligence algorithm is adopted, the intelligence level and the refinement degree of sheep house environment regulation and control are effectively improved, breakthrough is realized in the aspects of in-house temperature and humidity distribution uniformity, equipment response rate and regulation and control measure pertinence, and a new thought is provided for large-scale sheep farm environment management.

Description

Technical Field

[0001] The present invention relates to the technical field of animal husbandry environment control, and in particular to an intelligent temperature and humidity control system for a sheep house based on artificial intelligence. Background Art

[0002] As modern animal husbandry develops towards scale and intensification, the role of sheep house environmental control in large-scale farms has become increasingly prominent. As temperature and humidity sensitive livestock, the growth, development, immune function, and production performance of sheep are closely related to the temperature and humidity environment in the house. Ensuring that the temperature and humidity of the sheep house are stable within an appropriate range for a long time is a key measure to improve the health level of sheep and increase breeding efficiency.

[0003] At present, the temperature and humidity control of sheep houses generally adopts centralized air-conditioning systems, constant temperature and humidity controllers and other solutions. However, in practical applications, the following problems still need to be solved urgently: Affected by the layout of the sheep house and the distribution of equipment, the temperature and humidity levels of different functional areas are significantly different. The centralized control system cannot take into account the special needs of the local area environment; the control response speed of the existing temperature control equipment is slow, and it is difficult to make real-time feedback on the dynamic fluctuations of the environment in the sheep house. The adjustment is often triggered only after the temperature and humidity exceed the standard, which affects the growth experience of the sheep; sheep of different breeds and different growth stages have different requirements for the suitable range of temperature and humidity in the house. It is difficult for traditional environmental control systems to be tailored to the sheep, and the control measures are not targeted enough; traditional environmental control systems are mostly fixed installations, with low scalability and modularity, which is not conducive to flexible upgrades and renovations of farms according to actual needs; relying on manual setting of temperature and humidity thresholds and judgment of environmental abnormalities, mistakes are prone to occur. At the same time, frequent manual operations also increase labor intensity and affect management efficiency.

[0004] In view of the above problems, it is urgent to develop an intelligent control system for sheep house environment with stronger adaptability, faster response speed and more humanized functional design to better meet the needs of standardized and large-scale sheep production. The proposal of the present invention is an innovative research carried out to address the shortcomings of the above traditional technologies. Summary of the invention

[0005] The purpose of the present invention is to propose an intelligent control system that integrates deep reinforcement learning algorithm and zoning control strategy to address the common problems of weak zoning control ability, dynamic adjustment lag, lack of pertinence, etc. in the existing sheep house temperature and humidity control systems.

[0006] In order to achieve the above-mentioned invention object, the present invention adopts the following technical scheme: An artificial intelligence-based intelligent temperature and humidity control system for a sheep house, comprising a plurality of temperature and humidity sensors arranged in different functional areas in the sheep house, a data acquisition unit, a sheep house temperature and humidity adjustment execution device, and an artificial intelligence control unit.

[0007] Among them, the temperature and humidity sensor adopts a high-precision digital sensor, which is connected to the data acquisition unit through a bus communication, and collects temperature and humidity data of different areas in the sheep house at a fixed frequency.

[0008] The data acquisition unit adopts an industrial-grade microcomputer or an embedded microcontroller, collects data from each temperature and humidity sensor in a polling manner through a bus, and transmits the data to the artificial intelligence control unit.

[0009] The temperature and humidity control actuator for the sheep house includes an air conditioner, a humidifier, an exhaust fan, and an actuator responsible for controlling the working parameters of the above equipment. The air conditioner, humidifier, and exhaust fan have the functions of cooling, humidification, and ventilation, respectively, and are reasonably configured in layout and specifications according to the area and height of the sheep house. Each device is equipped with an actuator, which communicates with the artificial intelligence control unit through the field bus. The actuator integrates a controller and a drive circuit inside, and independently adjusts the working state of the corresponding device according to the received control instructions.

[0010] The artificial intelligence control unit is the core of the system, including a processor with high-performance computing capabilities, large-capacity memory, solid-state hard disk, and installed machine learning framework and algorithm program. Its workflow is as follows: Receive the temperature and humidity data of the sheep house partitions transmitted by the data acquisition unit as the state input of the deep reinforcement learning model; Build an independent deep reinforcement learning model for each partition, discretize the temperature and humidity adjustment range into a limited action space, and set a reward function that balances stability and energy saving; The action-value function is fitted using a deep neural network, and the state-action-reward data stream generated by continuous interaction with the sheep house environment is trained to make the neural network approach the optimal control strategy. Regularly receive changes in sheep's physiological parameters and dynamically adjust reward function parameters to enable control decisions to adapt to the needs of sheep at different growth stages; The action combination with the largest control action value calculated by the deep reinforcement learning model is converted into control instructions and transmitted to the actuators of the corresponding partitions through the field bus to achieve independent control of the partitions.

[0011] The control process of the present invention is as follows: the temperature and humidity sensor continuously collects environmental parameters of each partition in the sheep house → the data collection unit summarizes multi-point data and transmits it to the artificial intelligence control unit → the artificial intelligence control unit reads the environmental status of the partition and inputs it into the deep reinforcement learning model → the model calculates the expected returns of various control action combinations, and selects the combination with the largest return → sends control instructions to the actuator → the actuator independently controls the temperature and humidity adjustment equipment corresponding to the partition → receives feedback at the new sampling time, and calculates the reward value according to the target temperature and humidity range and the comfort of the sheep → uses the temporal difference algorithm to update the deep neural network parameters, and cyclically optimizes the control strategy.

[0012] Arrange multi-point sensing and independent control execution devices at the hardware level to meet the needs of fine environmental control for different functional areas of the sheep house; Deep reinforcement learning technology is introduced at the software level. Through continuous learning and strategy iteration, the system has the ability to dynamically perceive and quickly adjust the environment, similar to that of expert breeders, making the control process more intelligent and efficient. The sheep's physiological indicators are cleverly incorporated into the reward function of the reinforcement learning model, so that the optimization target changes dynamically with the growth stage of the sheep, and the environmental adaptability is greatly enhanced; The system architecture of distributed data acquisition and independent control of partitions allows local environmental control to be carried out without disturbing the overall situation, and the system has better flexibility and scalability.

[0013] To sum up, the intelligent temperature and humidity control system for sheep houses based on artificial intelligence proposed in the present invention has well solved the common shortcomings of traditional control systems, especially in terms of multi-partition independent control, adaptive optimization strategy learning, and dynamic response to sheep growth needs. It has achieved the overall energy-saving, high-efficiency, precise and intelligent technical effects, and provided a more advanced and practical new way for environmental control in modern large-scale sheep farms, and has broad prospects for promotion and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0015] Figure 1 This is a schematic diagram of the architecture of the intelligent temperature and humidity control system of the sheep house of the present invention; Figure 2 It is a schematic diagram of the physical layout of the sheep house of the present invention; Figure 3 It is a schematic diagram of the system data flow of the present invention; Figure 4 It is a schematic diagram of the DQN model structure of the present invention; Figure 5 It is a schematic diagram of the system control flow of the present invention; Figure 6 It is a schematic diagram of the partition control strategy of the present invention; Figure 7 A schematic diagram of the reward function design of the present invention. DETAILED DESCRIPTION

[0016] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0017] It should be noted that the references to "one embodiment", "embodiment", "exemplary embodiments", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0018] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0019] like Figure 1 As shown, an artificial intelligence-based temperature and humidity intelligent control system for a sheep house includes temperature and humidity sensors arranged in different areas of the sheep house, a data acquisition unit connected to the temperature and humidity sensors, a sheep house temperature and humidity control device, and an artificial intelligence control unit. The temperature and humidity control device includes an air conditioner, a humidifier, an exhaust fan, and an actuator connected to the control thereof. The artificial intelligence control unit is connected to the data acquisition unit and the actuator to receive the temperature and humidity data collected by the data acquisition unit, run a deep reinforcement learning algorithm, and output an adjustment instruction to the actuator, thereby controlling the air conditioner, humidifier, and exhaust fan to work and adjust the temperature and humidity in the sheep house.

[0020] Specifically, in this embodiment, the temperature and humidity sensors are arranged at representative positions of different partitions in the sheep house, and are used to collect temperature and humidity parameters of each area in real time. The data acquisition unit collects data from the temperature and humidity sensors at a preset frequency and transmits it to the artificial intelligence control unit.

[0021] The artificial intelligence control unit specifically uses the DeepQNetwork (DQN) algorithm in deep reinforcement learning to control the temperature and humidity of the sheep house. The real-time temperature and humidity status in the sheep house is used as input, and the reward value that may be obtained by taking various control actions (adjusting the air conditioning temperature, adjusting the air volume, turning on humidification) in the current state is calculated through the neural network, and the action with the largest reward value is selected, and the corresponding adjustment instruction is sent to the actuator. The actuator controls the working parameters of the air conditioner and humidifier equipment according to the adjustment instruction. After the control is completed, the artificial intelligence control unit updates the neural network parameters according to the control effect and the set objective function, so that the prediction of the reward value is more accurate and the control strategy is more reasonable.

[0022] When training the DQN model, the historical data collected by the temperature and humidity sensors in each area are used as samples, and the physiological parameters of the sheep in the same period, such as feed intake and rumination time, are integrated to set the reward function. When the temperature and humidity are close to the set comfort range and the sheep perform well, positive rewards are given; when the temperature and humidity deviate from the comfort range, or the sheep's feed intake decreases, negative rewards are given. Through a large number of iterations, the model learns to dynamically adjust the control strategy according to the status of the sheep house and the performance of the sheep.

[0023] The artificial intelligence control unit realizes independent regulation of different areas by means of zoning control. On the basis that each zone is independently equipped with temperature and humidity regulation equipment, the artificial intelligence control unit establishes a separate regulation model for each zone. Each zone model adopts the same DQN structure and training process, but sets different initial parameters and reward function parameters according to the differences in zone location, area, sheep breed, number, and age, so that the model can learn adaptive regulation strategies. In the actual control process, the artificial intelligence control unit runs each zone model in real time, independently controls the temperature and humidity regulation equipment in the zone, but at the same time considers the environmental parameters of adjacent areas to avoid excessive temperature and humidity differences between zones.

[0024] In short, the system uses a combination of distributed sensing, partition control, and deep reinforcement learning to accurately respond to environmental differences in various areas within the sheep house, adapt to the growth needs of sheep at different stages, dynamically optimize the control strategy, and achieve refined control of temperature and humidity in the sheep house, providing a more comfortable and healthy growth environment for the sheep. Those skilled in the art can understand that the partitions, number of sensors, type of control equipment, and neural network structure here can be adjusted according to the actual scale and needs of the sheep house to achieve the best control effect. In addition, the system can also be expanded to access other physiological parameters such as the surface temperature and weight of the sheep to further improve the pertinence of the control. Example

[0025] like Figure 2 As shown, an artificial intelligence-based intelligent temperature and humidity control system for sheep houses of this embodiment includes: The nine temperature and humidity sensors arranged in the sheep house are located in the representative positions of the northeast, north, northwest, east, center, west, southeast, south and southwest of the sheep house. The temperature and humidity sensor adopts SHT30 digital temperature and humidity sensor with a range of -40°C-125°C and an accuracy of ±0.3°C; 0100%RH, an accuracy of ±2%RH.

[0026] 1 data acquisition unit, using STM32F407 microcontroller and RS485 bus, collects data from 9 temperature and humidity sensors in sequence through the bus with a sampling frequency of 1 time / 5 minutes, and transmits the data to the artificial intelligence control unit.

[0027] The temperature and humidity control equipment of the sheep house includes 6 5-horsepower cabinet air conditioners, 4 humidifiers, and 6 exhaust fans, which are arranged around the sheep house. The air conditioner adopts Midea KFR-120QW / BP3DN1Y-DA400 (B2) variable frequency cabinet, with a rated cooling capacity of 12000W and an air volume of 1600m³ / h; the humidifier adopts Xinke SC-15Y130 ultrasonic industrial humidifier, with a humidification capacity of 15kg / h and atomization particle size of 1-5μm; the exhaust fan adopts Qilisu BPT-300 side wall fan, with a rated air volume of 8500m³ / h and a single power of 300W.

[0028] The artificial intelligence control unit uses 2 Intel Xeon Gold 6346 processors and 1 NVIDIA A100 Tensor Core GPU, 256GB of memory, 4TB SSD storage, and is installed with Ubuntu 20.04 operating system and TensorFlow 2.8 machine learning framework.

[0029] The 18 actuators are connected to the control ends of 6 air conditioners, 6 exhaust fans and 4 humidifiers respectively, receive control instructions from the artificial intelligence control unit, and adjust the working parameters of each device.

[0030] like Figure 5 as well as Figure 3 As shown, the control flow of the system is as follows: The artificial intelligence control unit receives the temperature and humidity data of 9 areas of the sheep house from the data acquisition unit.

[0031] The artificial intelligence control unit divides the sheep house into 4 control zones, each of which contains 2-3 temperature and humidity sensors. Figure 1 The numbers Ⅰ-Ⅳ marked in the dotted box are used to distinguish them.

[0032] like Figure 4As shown in the figure, for each partition, the AI ​​control unit runs an independent DQN model. Taking the Northeast Zone (Zone I) as an example, its state quantity is the temperature, humidity, and current time collected by the two temperature and humidity sensors in the zone, as well as the temperature and humidity collected by the temperature and humidity sensors in Zone IV (as the state of the adjacent zone); the action space is the air outlet mode (cooling, heating, air supply), set temperature (16-30°C), air volume (low, medium, high) of an air conditioner in Zone I, the switches (on and off) of two exhaust fans and one humidifier, a total of 108 action combinations; the reward function is set to the highest reward value when the temperature of the sheep house is close to 26°C and the humidity is close to 60%, and a negative reward item is superimposed to punish frequent adjustment behaviors. The model uses CNN to extract state features, and connects the fully connected layer and Duelinghead to output the Q value of each action.

[0033] like Figure 4 As shown in the figure, the DQN model predicts the Q value of various control action combinations according to the status of each area of ​​the sheep house, selects the action with the highest Q value, and outputs control instructions to the actuators in the corresponding partitions. The actuators adjust the working parameters of the air conditioner, exhaust fan, and humidifier according to the instructions.

[0034] At the next sampling moment, the AI ​​control unit receives the regulated temperature and humidity feedback, calculates the reward value based on the degree of proximity, and uses the time difference algorithm to update the neural network parameters. This process is repeated continuously to enable the model to continuously learn and optimize.

[0035] like Figure 5 As shown, in actual application, the artificial intelligence control unit runs the above control process every 5 minutes to dynamically control the temperature and humidity environment in the sheep house in real time.

[0036] like Figure 6 as well as Figure 7 As shown in the figure, in addition, considering that sheep at different growth stages have different requirements for environmental parameters, the system also introduces an external interface for manually inputting information about the breed, age, and quantity of sheep. The artificial intelligence control unit automatically adjusts the reward function of each partitioned DQN model based on the input parameters, so that the lambing model determines the environment with a temperature of 26-28°C and a humidity of 65% as the optimal, while the fattening model determines the environment with a temperature of 16-22°C and a humidity of 55-65% as the optimal.

[0037] After 30 days of training and testing, the system of this embodiment showed that the average daily fluctuation of temperature in each zone was less than ±1.5°C, and the average daily fluctuation of humidity was less than ±5%RH. The sheep's feed intake increased by 4.7%, and their weight increased by 8.3%, with a significant improvement in the overall effect.

[0038] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0039] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An intelligent temperature and humidity control system for sheep houses based on artificial intelligence, characterized in that , including multiple temperature and humidity sensors, data acquisition units, sheep house temperature and humidity adjustment execution devices, and artificial intelligence control units arranged in different functional areas in the sheep house; the temperature and humidity sensors use digital sensors and are connected to the data acquisition unit through a bus; the data acquisition unit uses an industrial-grade computer or an embedded controller to collect data from each temperature and humidity sensor at a fixed frequency and transmit it to the artificial intelligence control unit; the sheep house temperature and humidity adjustment execution device includes an air conditioner, a humidifier, and an exhaust fan, and their working parameters are controlled by actuators equipped with controllers and drive circuits respectively, and the actuators receive control instructions from the artificial intelligence control unit through a field bus; the artificial intelligence control unit contains a processor, memory, storage medium, machine learning framework, and algorithm program, which calculates the optimal temperature and humidity adjustment action combination for the sheep house partition through a deep reinforcement learning model, and sends the control instructions to the adjustment execution device of the corresponding partition.

2. The system according to claim 1, characterized in that The deep reinforcement learning model takes the partitioned environment state as input and the range of equipment parameter changes as the action space. It introduces a dynamic reward function based on the physiological indicators of sheep, fits the state-action value function through a deep neural network, and continuously trains using the sheep house feedback data stream based on the temporal difference algorithm, so that the output action combination can be adaptively adjusted as the sheep grows.

3. The system as claimed in claim 1, characterized in that The artificial intelligence control unit builds an independent deep reinforcement learning model for each functional partition of the sheep house. The partition division comprehensively considers spatial distance, environmental differences, and functional factors. Each partition is equipped with independently controlled temperature and humidity adjustment equipment.

4. The system according to claim 1, characterized in that The layout density of the temperature and humidity sensors is not less than 1 per 10-20 square meters, and the sampling frequency is not less than 1 per 5-10 minutes.

5. The system according to claim 1, characterized in that ,The data acquisition unit has the functions of data preprocessing, exception elimination, and data compression, and the MQTT protocol is used for data transmission between ,the artificial intelligence control unit.

6. The system as claimed in claim 1, characterized in that The air conditioner adopts variable frequency air conditioner, and the horsepower is determined according to the partition area and volume; the humidifier adopts ultrasonic industrial humidifier, and the humidification capacity is configured according to the size of the partition space; the exhaust fan adopts centrifugal fan or side wall fan whose air volume matches the partition volume, and the power meets the requirements of ventilation times per unit time.

7. The system as claimed in claim 2, characterized in that ,The deep neural network model uses convolutional neural network to extract ,temporal features of temperature and humidity, and LSTM is used in the middle layer to capture ,the temporal association of environmental changes, and the ,strategy network and value network independently output the probability and ,state value of each action, respectively. The overall network scale is in the hundreds of thousands of ,parameters, which can be trained and converged by gradient descent method.

8. The system as claimed in claim 2, characterized in that ,The setting of the dynamic reward function needs to take into account both environmental stability and energy saving. The setting of the standard temperature and humidity range needs to refer to the sheep breed, growth stage, and physiological condition. When the actual environmental parameters deviate from the target range, the negative reward value is calculated according to the degree of deviation. At the same time, additional penalties are imposed on the frequent starting and stopping of the temperature and humidity control equipment.

9. A method for controlling temperature and humidity in a sheep house based on the system according to any one of claims 1 to 8, characterized in that , including the following steps: S1. Install temperature and humidity sensors to collect environmental parameters of different functional areas in the sheep house; S2, collect multi-point sensor data through the data acquisition unit and transmit it to the artificial intelligence control unit at a fixed frequency; S3, the AI ​​control unit reads the state of the partition environment, inputs the deep reinforcement learning model, calculates the expected returns of various control action combinations, and selects the action combination with the maximum return; S4, sending control instructions to the actuator corresponding to the partition, and independently adjusting the temperature and humidity control equipment of the partition; S5. At the next sampling moment, the feedback reward value is calculated according to the new environmental parameters and the sheep's physiological indicators, and the deep neural network parameters are updated using the temporal difference algorithm; S6. Repeat iterative steps (3)-(5) to continuously optimize the environmental control strategy.

10. The method according to claim 9, characterized in that In step (1), the density of temperature and humidity sensors shall not be less than one per 10-20 square meters; in step (2), the frequency of data transmission shall not be less than once every 5-10 minutes; in step (3), the action and reward function settings in the deep reinforcement learning model shall meet the limitations of claims 2 and 8.

Images