Greenhouse control method, device and equipment based on artificial intelligence and storage medium

Abstract

The application provides a greenhouse control method and device based on artificial intelligence, electronic equipment and computer readable storage medium, and relates to the technical field of artificial intelligence. The method comprises the following steps: acquiring the growth state of plants in a greenhouse in a first period, the internal climate state of the greenhouse in the first period, and the external weather state of the greenhouse in the first period; calling a machine learning model based on the growth state in the first period, the internal climate state in the first period, and the external weather state in the first period to obtain environment control information for controlling the greenhouse in a second period; and applying the environment control information to the greenhouse in the second period. Through the application, intelligent control of plant growth in the greenhouse can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a greenhouse control method and device based on artificial intelligence, an electronic device, and a computer readable storage medium. BACKGROUND

[0002] Artificial intelligence (AI) is a comprehensive technology of computer science, which enables machines to have the functions of perception, reasoning and decision-making by studying the design principles and implementation methods of various intelligent machines. Artificial intelligence technology is a comprehensive discipline, involving a wide range of fields, such as natural language processing technology and machine learning / deep learning. With the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0003] In the related art, the growth environment of plants in a greenhouse is adjusted by manual control based on the experience of planting experts to control the growth state of the plants. There is a lack of effective solutions for adjusting the growth environment of plants in a greenhouse based on artificial intelligence. SUMMARY

[0004] The embodiments of the present application provide a greenhouse control method and device based on artificial intelligence, an electronic device, and a computer readable storage medium, which can realize intelligent control of plant growth in a greenhouse.

[0005] The technical solutions of the embodiments of the present application are as follows:

[0006] The embodiments of the present application provide a greenhouse control method based on artificial intelligence, comprising:

[0007] obtaining the growth state of plants in a greenhouse in a first period, the internal climate state of the greenhouse in the first period, and the external weather state of the greenhouse in the first period;

[0008] calling a machine learning model based on the growth state in the first period, the internal climate state in the first period, and the external weather state in the first period to obtain environment control information for controlling the greenhouse in a second period, wherein the second period is later than the first period;

[0009] applying the environment control information to the greenhouse in the second period.

[0010] The embodiments of the present application provide an image target recognition device, comprising:

[0011] an obtaining module configured to obtain the growth state of plants in a greenhouse in a first period, the internal climate state of the greenhouse in the first period, and the external weather state of the greenhouse in the first period;

[0012] a processing module configured to invoke a machine learning model based on the growth state of the first period, the internal climate state of the first period, and the external weather state of the first period to obtain environment control information for controlling the greenhouse in a second period, wherein the second period is later than the first period;

[0013] an application module configured to apply the environment control information to the greenhouse in the second period.

[0014] In the technical solution described above, the processing module is further configured to perform the following processing based on the machine learning model:

[0015] determine an expected growth state in the second period that meets a planting target based on the growth state of the first period, and determine an internal climate state of the second period when the expected growth state is achieved;

[0016] determine the environment control information for controlling the greenhouse in the second period based on the characteristics of the internal climate state of the second period being jointly affected by the external weather state of the second period and the environment control information for controlling the greenhouse in the second period.

[0017] In the technical solution described above, the machine learning model comprises a first mapping network and a fusion network; and the processing module is further configured to perform mapping processing on the growth state of the first period based on the first mapping network to obtain the expected growth state in the second period that meets the planting target.

[0018] map the internal climate state of the second period to the environment control information for controlling the greenhouse in the second period based on the mapping relationship between the environment control information and the external weather state of the greenhouse and the internal climate state of the greenhouse included in the fusion network.

[0019] In the technical solution described above, the machine learning model further comprises a second mapping network; and the processing module is further configured to perform mapping processing on the expected growth state of the second period based on the mapping relationship between the internal climate state of the greenhouse and the growth state of the plant included in the second mapping network to obtain the internal climate state of the second period when the expected growth state of the second period is achieved; or

[0020] perform state conversion processing on the state difference between the growth state of the first period and the expected growth state of the second period based on the mapping relationship between the internal climate state of the greenhouse and the change in the growth state of the plant included in the second mapping network to obtain the internal climate state of the second period when the expected growth state of the second period is achieved.

[0021] In the technical solution, the fusion network comprises a first convolutional layer, a second convolutional layer, a fully connected layer, and a third convolutional layer; the processing module is further configured to perform convolutional processing on the external weather state of the second period based on the first convolutional layer to obtain first state information corresponding to the external weather state;

[0022] perform convolutional processing on the internal climate state of the second period based on the second convolutional layer to obtain second state information corresponding to the internal climate state;

[0023] determine a difference between the second state information and the first state information based on the fully connected layer;

[0024] perform convolutional processing on the difference based on the third convolutional layer to obtain environmental control information for controlling the greenhouse in the second period.

[0025] In the technical solution, the device further comprises:

[0026] a training module configured to construct a training sample of the machine learning model based on the growth state of plants in the greenhouse in a plurality of historical periods, the internal climate state of the greenhouse in the plurality of historical periods, and the external weather state of the greenhouse in the plurality of historical periods;

[0027] train the machine learning model based on the training sample to obtain the machine learning model for predicting environmental control information.

[0028] In the technical solution, the training module is further configured to perform the following processing for any historical period in the plurality of historical periods:

[0029] obtain the growth state of a next historical period of the historical period, the internal climate state of the next historical period of the historical period, and the external weather state of the next historical period of the historical period;

[0030] combine the growth state of plants in the greenhouse in the historical period, the internal climate state of the greenhouse in the historical period, and the external weather state of the historical period into first state information of the historical period;

[0031] combine the growth state of the next historical period of the historical period, the internal climate state of the next historical period of the historical period, and the external weather state of the next historical period of the historical period into second state information of the next historical period;

[0032] construct a training sample of the historical period based on the first state information of the historical period, environmental control information for controlling the greenhouse in the historical period, and the second state information of the next historical period.

[0033] combine the training samples of the plurality of historical periods to obtain the training sample of the machine learning model.

[0034] The device further comprises:

[0035] The storage module is configured to store the training samples of the historical periods in a cache space.

[0036] The training module is further configured to obtain the training samples of the plurality of historical periods from the cache space when the number of the training samples of the historical periods in the cache space reaches a set threshold, and train the machine learning model based on the training samples of the plurality of historical periods.

[0037] In the technical solution, the training module is further configured to construct a target function of the machine learning model based on the training sample of any historical period in the training sample and the labeled evaluation parameter of the historical period.

[0038] The parameters of the machine learning model are updated until the target function converges, and the updated parameters of the machine learning model when the target function converges are used as the parameters of the machine learning model for predicting the environmental control information.

[0039] The processing module is further configured to call the machine learning model based on the training sample of any historical period in the training sample to perform prediction processing, and obtain predicted environmental control information for controlling the greenhouse in the historical period.

[0040] Based on the predicted environmental control information, a predicted evaluation parameter of the historical period is obtained.

[0041] The training module is further configured to construct a target function of the machine learning model based on the training sample of the historical period, the predicted evaluation parameter of the historical period, and the labeled evaluation parameter of the historical period.

[0042] The application module is further configured to obtain a growth state of a next historical period in the training sample of the historical period.

[0043] A plant simulator model is called based on the growth state of the next historical period to determine growth expectation information caused by the plant growth in the historical period, and the growth expectation information is used as the labeled evaluation parameter of the historical period.

[0044] The application module is further configured to obtain environmental control information for controlling the greenhouse in the historical period.

[0045] The plant simulator model is invoked based on the environmental control information used to control the greenhouse during the historical period to determine the resource information required by the environmental control information.

[0046] The difference between the growth expectation information and the resource information is used as the annotation and evaluation parameter for the historical period.

[0047] This application provides an electronic device for image target recognition, the electronic device comprising:

[0048] Memory, used to store executable instructions;

[0049] The processor, when executing executable instructions stored in the memory, implements the image target recognition method provided in the embodiments of this application.

[0050] This application provides a computer-readable storage medium storing executable instructions, which, when executed by a processor, implement the image target recognition method provided in this application.

[0051] The embodiments of this application have the following beneficial effects:

[0052] By calling a machine learning model, based on the growth status, internal climate status, and external weather status of the first period, environmental control information for controlling the greenhouse in the second period is obtained. This information is then combined with the external weather status to dynamically adjust and control the growth environment of the plants in the greenhouse, thereby achieving intelligent and precise control of plant growth. Attached Figure Description

[0053] Figure 1 This is a schematic diagram illustrating an application scenario of the greenhouse control system provided in this application embodiment;

[0054] Figure 2 This is a schematic diagram of the structure of an electronic device for greenhouse control provided in an embodiment of this application;

[0055] Figures 3-5 This is a flowchart illustrating the artificial intelligence-based greenhouse control method provided in an embodiment of this application;

[0056] Figure 6 This is a flowchart illustrating the artificial intelligence-based greenhouse control method provided in an embodiment of this application;

[0057] Figure 7 This is a schematic diagram of the structure of the machine learning model provided in the embodiments of this application;

[0058] Figure 8 This is a schematic diagram of the platform architecture provided in the embodiments of this application;

[0059] Figure 9This is a schematic diagram of crop planting simulation provided in the embodiments of this application;

[0060] Figure 10 This is a schematic diagram of the workflow of the data collection module provided in the embodiments of this application. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0062] In the following description, the terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first" and "second" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0063] It is understood that in the embodiments of this application, data such as user information are involved. When the embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with relevant laws, regulations and standards.

[0064] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0065] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0066] 1) Convolutional Neural Networks (CNNs): A class of feedforward neural networks (FNNs) that include convolutional computations and have a deep structure, CNNs are one of the representative algorithms of deep learning. CNNs possess representation learning capabilities, enabling them to perform shift-invariant classification of input images according to their hierarchical structure.

[0067] 2) Environmental control information: This is used to control environmental factors that affect plant growth, such as temperature settings, CO2 concentration settings, fluorescent light switching time, irrigation time, etc.

[0068] 3) Growth status: Used to characterize the stage of plant growth or the vigor of plant growth, such as plant height, leaf area index, and fruit weight.

[0069] 4) External weather conditions: used to characterize weather conditions outside the greenhouse, such as outdoor temperature and outdoor humidity.

[0070] 5) Internal climate conditions: used to characterize the climate conditions inside the greenhouse, such as the internal temperature and CO2 concentration.

[0071] Intelligent greenhouses enable precise control over crop cultivation, making the development of superior cultivation control strategies a crucial research area. The cultivation strategies employed in related technologies primarily fall into two categories: the first stems from the cultivation experience of agricultural experts, who combine past cultivation data with their professional knowledge to set rules for the cultivation process—this is currently the most widely adopted method in greenhouses; the second originates from data mining, using deep learning and other techniques to uncover the correlation between control strategies and cultivation benefits from past cultivation datasets, thereby obtaining the optimal cultivation strategy.

[0072] However, the planting strategies in related technologies have the following problems:

[0073] 1) Relying on planting strategies from agricultural experts is relatively expensive. In addition, the planting strategies proposed by agricultural experts are generally fixed logic strategies that cannot be dynamically changed as the planting process progresses, resulting in poor adaptability. Furthermore, the control granularity of planting strategies proposed by experts is also relatively coarse, generally consisting of combinations of logic, such as cooling down when the real-time temperature exceeds a certain temperature threshold. It is impossible to combine all state information for fine-grained control, such as the specific light intensity and temperature settings for each hour.

[0074] 2) Although data mining-based planting strategies do not require expert knowledge intervention and can directly establish a connection between greenhouse status data and planting result data during the planting process, this method requires a large amount of agricultural data for training, which is difficult to collect. Moreover, the collected agricultural planting data is generally based on certain fixed planting strategies with little variation, making it difficult to discover better strategies than planting experts. Furthermore, there is relatively little data about the crop itself in the agricultural planting data, making it difficult to consider the crop's own condition as an important factor in determining the next planting measures.

[0075] To address the aforementioned issues, this application provides an artificial intelligence-based greenhouse control method, apparatus, electronic device, and computer-readable storage medium, which enables intelligent control of plant growth in greenhouses.

[0076] The AI-based method for intelligently controlling the plant growth environment in a greenhouse, provided in this application embodiment, can be implemented independently by a terminal / server; or it can be implemented collaboratively by a terminal and a server. For example, the terminal can independently implement the AI-based method for intelligently controlling the plant growth environment in a greenhouse as described below, and apply environmental control information to the greenhouse in a second period through a control device. Alternatively, the terminal can send a control request for the greenhouse to the server (including the growth status of the plants in the greenhouse in the first period, the internal climate status of the greenhouse in the first period, and the external weather status of the greenhouse in the first period). The server executes the AI-based greenhouse control method according to the received greenhouse control request, and in response to the control request, applies environmental control information to the greenhouse in a second period through a control device to automatically control the growth of the plants in the greenhouse.

[0077] The electronic devices for greenhouse control provided in this application can be various types of terminal devices or servers. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, etc., but is not limited to these. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited herein.

[0078] Taking servers as an example, such as server clusters deployed in the cloud, AI as a Service (AIaaS) is offered to users. The AIaaS platform breaks down several common AI services and provides them as independent or packaged services in the cloud. This service model is similar to an AI-themed marketplace, where all users can access and use one or more AI services provided by the AIaaS platform through application programming interfaces.

[0079] For example, one type of AI cloud service could be a greenhouse control service, where a cloud server encapsulates the greenhouse control program provided in this application embodiment. Users invoke the greenhouse control service in the cloud service via a terminal (running a client, such as a greenhouse monitoring client), causing the cloud-deployed server to invoke the encapsulated greenhouse control program. Based on the growth status of plants in the greenhouse during a first period, the internal climate status of the greenhouse during the first period, and the external weather status of the greenhouse during the first period, environmental control information for controlling the greenhouse in a second period is determined. This environmental control information is then applied to the greenhouse during the second period to monitor plant growth. For example, in a greenhouse monitoring application, historical internal climate status is obtained through sensors in the greenhouse (such as temperature sensors and humidity sensors), and current external weather status is obtained through a third application. Based on historical internal climate status, historical growth status, and current weather forecast, environmental control information for controlling the greenhouse in the current period is determined. This environmental control information is then applied to the greenhouse during the second period through control equipment (such as humidifiers, heaters, and supplemental lighting) to automatically control plant growth in the greenhouse.

[0080] See Figure 1 , Figure 1 This is a schematic diagram of the application scenario of the greenhouse control system 10 provided in the embodiment of this application. The terminal 200 is connected to the server 100 through the network 300. The network 300 can be a wide area network or a local area network, or a combination of the two.

[0081] Terminal 200 (running a client, such as a greenhouse monitoring client) can be used to obtain control requests for the greenhouse. For example, when a preset time point is reached, i.e. the first period, terminal 200 obtains the external weather conditions of the first period and the external weather conditions of the second period (weather forecast or real-time weather data) through a third application. Terminal 200 obtains the internal climate conditions and growth conditions of the first period (historical period) through sensors in the greenhouse. After terminal 200 collects the internal climate conditions, external weather conditions and growth conditions of the first period, it automatically generates a control request for the greenhouse.

[0082] In some embodiments, a greenhouse control plugin may be embedded in the client running on the terminal to implement an AI-based greenhouse control method locally on the client. For example, after the terminal 200 obtains a control request for the greenhouse (including the internal climate state, the external weather state, and the growth state in the first period), it calls the greenhouse control plugin to implement an AI-based greenhouse control method. Based on the growth state of the plants in the greenhouse in the first period, the internal climate state of the greenhouse in the first period, and the external weather state of the greenhouse in the first period, it determines the environmental control information for controlling the greenhouse in the second period and applies the environmental control information to the greenhouse in the second period to monitor the growth of the plants through the environmental control information in the greenhouse.

[0083] In some embodiments, after receiving a control request for the greenhouse, the terminal 200 calls the greenhouse control interface of the server 100 (which can be provided as a cloud service, i.e., a greenhouse control service). The server 100 calls a machine learning model to determine environmental control information for controlling the greenhouse in a second period based on the growth status of the plants in the greenhouse in a first period, the internal climate status of the greenhouse in the first period, and the external weather status of the greenhouse in the first period. This environmental control information is then applied to the greenhouse in the second period to monitor plant growth. For example, in a greenhouse monitoring application, the terminal 200 obtains historical internal climate status and historical growth status through sensors in the greenhouse, and transmits this information via a third-party application. By acquiring historical and current weather forecasts, and based on historical internal climate conditions, growth conditions, and weather forecasts, a control request for the greenhouse is automatically generated and sent to server 100. Server 100 calls a machine learning model to determine the environmental control information for controlling the greenhouse in the current period, based on the historical growth conditions of the plants, the internal climate conditions of the greenhouse, and the external weather conditions of the greenhouse. The environmental control information for the current period is then sent to terminal 200. Terminal 200 applies the environmental control information to the greenhouse in the current period through control devices to automatically control the growth of the plants in the greenhouse.

[0084] The structure of the electronic device for greenhouse control provided in the embodiments of this application is described below. See also... Figure 2 , Figure 2 This is a schematic diagram of the structure of an electronic device 500 for greenhouse control provided in an embodiment of this application. The explanation will take a server as an example. Figure 2The illustrated electronic device 500 for greenhouse control includes at least one processor 510, a memory 550, at least one network interface 520, and a user interface 530. The various components in the electronic device 500 are coupled together via a bus system 540. It is understood that the bus system 540 is used to enable communication between these components. In addition to a data bus, the bus system 540 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 2 The general labeled all buses as Bus System 540.

[0085] The processor 510 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0086] Memory 550 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 550 described in this application embodiment is intended to include any suitable type of memory. Memory 550 may optionally include one or more storage devices physically located away from processor 510.

[0087] In some embodiments, memory 550 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0088] Operating system 551 includes system programs for handling various basic system services and performing hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., for implementing various basic business functions and handling hardware-based tasks;

[0089] The network communication module 552 is used to reach other computing devices via one or more (wired or wireless) network interfaces 520, exemplary network interfaces 520 including: Bluetooth, WiFi, and Universal Serial Bus (USB), etc.

[0090] In some embodiments, the greenhouse control device provided in this application can be implemented in software, for example, it can be a greenhouse control plug-in in the terminal mentioned above, or a greenhouse control service in the server mentioned above. Of course, it is not limited to this, and the greenhouse control device provided in this application can be provided in various software embodiments, including various forms such as applications, software, software modules, scripts or code.

[0091] Figure 2 A greenhouse control device 555 stored in a memory 550 is shown. It can be software in the form of programs and plug-ins, such as a greenhouse control plug-in, and includes a series of modules, including an acquisition module 5551, a processing module 5552, an application module 5553, a training module 5554, and a storage module 5555. The acquisition module 5551, the processing module 5552, the application module 5553, and the storage module 5555 are used to implement the greenhouse control functions provided in the embodiments of this application, and the training module 5554 is used to train a machine learning model.

[0092] As previously stated, the AI-based greenhouse control method provided in this application can be implemented by various types of electronic devices. See also Figure 3 , Figure 3 This is a flowchart illustrating the artificial intelligence-based greenhouse control method provided in this application embodiment, combined with... Figure 3 The steps shown are explained.

[0093] In the following steps, the external weather conditions for the first period can be real-time weather forecasts obtained through a third application, meaning the external weather conditions for the first period are the actual weather conditions for that period. Alternatively, they can be weather data from the same period in the past, meaning the external weather conditions for the first period are simulated weather conditions for that period. The first period is a historical period relative to the second period, and the units for both periods can be months, weeks, days, or even hours.

[0094] The internal climate status is obtained by sensing the growth environment in the greenhouse through sensors, and the growth status is obtained by sensing the growth of the plants in the greenhouse through sensors.

[0095] In step 101, the growth status of the plants in the greenhouse during the first period, the internal climate status of the greenhouse during the first period, and the external weather status of the greenhouse during the first period are obtained.

[0096] like Figure 6As shown, the terminal obtains the internal climate and growth status of the greenhouse relative to the current period (second period) and the historical period (first period) through sensors in the greenhouse. It obtains the external weather status of the first period and the external weather status of the second period through a third application. Based on the growth status, internal climate status, and external weather status of the first period, the terminal automatically generates a control request for the greenhouse and sends the control request to the server. After receiving the control request, the server parses the control request to obtain the growth status, internal climate status, and external weather status of the first period. This information is then used to call a machine learning model to dynamically optimize the environmental control information.

[0097] In step 102, a machine learning model is invoked based on the growth status, internal climate status, and external weather status of the first period to obtain environmental control information for controlling the greenhouse in the second period, wherein the second period is later than the first period.

[0098] For example, by inputting the growth status, internal climate status, and external weather status of the first period into a machine learning model, the machine learning model learns the following strategy: using the greenhouse environmental control information and external weather status together to apply to the internal climate status of the greenhouse, and applying the internal climate status of the greenhouse to the growth status of the plants. Through predictive processing by the machine learning model, the environmental control information used to control the greenhouse in the second period is obtained, which is the environmental control information used to control the greenhouse in the next period. Combined with the external weather status, the growth environment of the plants in the greenhouse is dynamically adjusted and controlled, thereby achieving intelligent and precise control of the plant growth environment.

[0099] See Figure 4 , Figure 4 This is an optional flowchart illustrating an artificial intelligence-based greenhouse control method provided in an embodiment of this application. Figure 4 Show Figure 3 Step 102 can be achieved through steps 1021-1023: the following processing is performed based on the machine learning model: in step 1021, based on the growth status of the first period, the expected growth status in the second period that meets the planting target is determined; in step 1022, the internal climate status of the second period when the expected growth status is achieved is determined; in step 1023, based on the characteristics of the combined effect of the external weather status of the second period and the environmental control information used to control the greenhouse in the second period on the internal climate status of the second period, the environmental control information used to control the greenhouse in the second period is determined.

[0100] For example, after the server receives a control request for the greenhouse from the terminal, it parses the request to obtain the growth status, internal climate status, and external weather status of the first period. It then calls a machine learning model for predictive processing. Based on the strategy learned by the machine learning model, it first predicts the growth status based on the growth status of the first period to obtain the expected growth status for the second period that meets the planting objectives. The planting objectives can be pre-set according to the actual scenario, such as maximizing plant yield or maximizing the profit generated by the plants. After obtaining the expected growth status for the second period, it determines the internal climate status of the second period to achieve this expected growth status. Finally, based on the characteristics of the interaction between the external weather status and the environmental control information used to control the greenhouse in the second period, as learned by the machine learning model, it determines the environmental control information used to control the greenhouse in the second period, enabling intelligent control of the plant growth environment within the greenhouse.

[0101] As an example, the first growth stage is characterized by a fruit maturity of 10%. Based on the goal of maximizing profits from the plant, the growth stage is predicted based on the first stage, resulting in a predicted fruit maturity of 15% for the second stage. To achieve this, the internal climate conditions for the second stage are a greenhouse temperature of 28°C and a carbon dioxide concentration of 300 g / m³. The external weather conditions for the second stage are obtained from a third-party weather forecast, with an outdoor temperature of 20°C. Based on the characteristics of the interaction between the external weather conditions and the environmental control information used to control the greenhouse in the second stage on the internal climate conditions, the environmental control information for the second stage is determined as follows: temperature is set to 28°C, CO2 concentration is set to 300 g / m³, and fluorescent light switching time is set to 3 hours.

[0102] In some embodiments, the machine learning model includes a first mapping network, a second mapping network, and a fusion network; the growth state in the first period is mapped based on the first mapping network to obtain the expected growth state in the second period that meets the planting objectives; in contrast, the expected growth state in the second period is mapped based on the mapping relationship between the internal climate state of the greenhouse and the growth state of the plants included in the second mapping network to obtain the internal climate state of the second period when the expected growth state of the second period is achieved; in contrast, the internal climate state in the second period is mapped to environmental control information used to control the greenhouse in the second period based on the mapping relationship between the environmental control information of the greenhouse and the external weather conditions included in the fusion network and the internal climate state of the greenhouse.

[0103] Following the example above, such as Figure 7As shown, the growth state in the first period is input into the first mapping network of the machine learning model. The first mapping network maps the growth state in the first period to obtain the expected growth state in the second period that meets the planting objectives. This expected growth state is then input into the second mapping network. Through the mapping relationship between the internal climate state of the greenhouse and the plant growth state learned by the second mapping network, the expected growth state in the second period is mapped to obtain the internal climate state in the second period when the expected growth state is achieved. This internal climate state in the second period is then input into a fusion network. Through the mapping relationship between the greenhouse's environmental control information and external weather conditions and the greenhouse's internal climate state learned by the fusion network, the internal climate state in the second period is mapped to the environmental control information used to control the greenhouse in the second period. Therefore, the environmental control information for the second period can be predicted through multiple layers of networks in the machine learning model, thereby achieving intelligent and precise control of the plant growth environment and avoiding manual adjustments to the plant growth environment in the greenhouse.

[0104] In some embodiments, the machine learning model includes a first mapping network, a second mapping network, and a fusion network; the growth state in the first period is mapped based on the first mapping network to obtain the expected growth state in the second period that meets the planting objectives; in contrast, based on the mapping relationship between the internal climate state of the greenhouse and the changes in the plant growth state included in the second mapping network, the state difference between the growth state in the first period and the expected growth state in the second period is processed to obtain the internal climate state in the second period when the expected growth state in the second period is achieved; in contrast, based on the mapping relationship between the environmental control information of the greenhouse and the external weather conditions included in the fusion network and the internal climate state of the greenhouse, the internal climate state in the second period is mapped to the environmental control information used to control the greenhouse in the second period.

[0105] like Figure 7As shown, the growth state in the first period is input into the first mapping network of the machine learning model. The first mapping network maps the growth state in the first period to obtain the expected growth state in the second period that meets the planting objectives. This expected growth state is then input into the second mapping network. Through the mapping relationship between the internal climate state of the greenhouse and the changes in plant growth state learned by the second mapping network, the state difference between the growth state in the first period and the expected growth state in the second period is processed to obtain the internal climate state in the second period when the expected growth state is achieved. The second mapping network inputs the internal climate state of the second period into a fusion network. Through the mapping relationship between the greenhouse environmental control information and external weather conditions and the internal climate state learned by the fusion network, the internal climate state of the second period is mapped to the environmental control information used to control the greenhouse in the second period. Therefore, the environmental control information for the second period can be predicted through multiple layers of networks in the machine learning model, thereby achieving intelligent and precise control of the plant growth environment and avoiding manual adjustments to the plant growth environment in the greenhouse.

[0106] Continuing with the example above, the first growth stage is characterized by a fruit maturity of 10%, and the expected growth stage in the second stage is a fruit maturity of 15%. Therefore, the change in plant growth stage is a 5% increase in fruit maturity. To achieve this change, the internal climate conditions in the second stage are a greenhouse temperature of 28°C and a carbon dioxide concentration of 300 g / m³. Through the second mapping network, the relationship between the internal climate conditions of the greenhouse and the changes in plant growth stage during actual plant growth is learned. This allows for the accurate determination of the internal climate conditions in the second stage to achieve the expected growth stage, enabling the accurate location of environmental control information for controlling the greenhouse during the second stage.

[0107] In some embodiments, the fusion network includes a first convolutional layer, a second convolutional layer, a fully connected layer, and a third convolutional layer; mapping the internal climate state of the second period to environmental control information for controlling the greenhouse in the second period includes: performing convolution processing on the external weather state of the second period based on the first convolutional layer to obtain first state information corresponding to the external weather state; performing convolution processing on the internal climate state of the second period based on the second convolutional layer to obtain second state information corresponding to the internal climate state; determining the difference between the second state information and the first state information based on the fully connected layer; and performing convolution processing on the difference based on the third convolutional layer to obtain environmental control information for controlling the greenhouse in the second period.

[0108] like Figure 7As shown, the second mapping network inputs the internal climate state of the second period into the fusion network. Based on the learned mapping relationship between the greenhouse environmental control information and the external weather state and the internal climate state of the greenhouse, the external weather state of the second period is convolved by the first convolutional layer in the fusion network to obtain the first state information (i.e., the first convolutional vector) corresponding to the external weather state. Then, the internal climate state of the second period is convolved by the second convolutional layer in the fusion network to obtain the second state information (i.e., the second convolutional vector) corresponding to the internal climate state. Then, the difference between the second state information and the first state information is calculated by the fully connected layer to obtain the difference value between the second state information and the first state information. Finally, the difference value is convolved by the third convolutional layer to obtain the environmental control information used to control the greenhouse in the second period.

[0109] In step 103, environmental control information is applied to the greenhouse in the second period.

[0110] For example, the terminal sends a control request for the greenhouse to the server. After receiving the control request, the server calls a machine learning model to perform predictive processing, obtains the environmental control information for the second period, and feeds the environmental control information of the second period back to the terminal. The terminal then applies the environmental control information to the greenhouse in the second period through the control equipment, thereby transforming the internal climate state of the first period to the internal climate state of the second period. Under the combined effect of the internal climate state and the external weather state of the second period, the growth state of the first period is transformed to the growth state of the second period, thus realizing the automated control of plant growth in the greenhouse.

[0111] See Figure 5 , Figure 5 This is an optional flowchart illustrating an artificial intelligence-based greenhouse control method provided in an embodiment of this application. Figure 5 This demonstrates training methods for machine learning models. Figure 5 Show Figure 3 It also includes steps 104-105: In step 104, training samples for the machine learning model are constructed based on the growth status of plants in the greenhouse in multiple historical periods, the internal climate status of the greenhouse in multiple historical periods, and the external weather status of the greenhouse in multiple historical periods; In step 105, the machine learning model is trained based on the training samples to obtain a machine learning model for predicting environmental control information.

[0112] For example, the current state (including growth status, internal climate status, and external weather status) can be used as the state of the next historical period. The state of a plant within a growth cycle in a greenhouse (including growth status, internal climate status, and external weather status) can be used as training samples to train a machine learning model. That is, multiple historical periods belong to a plant growth cycle. For example, if the plant growth cycle is one month and the environmental control information is updated on a daily basis, then the number of multiple historical periods is 30, namely day 1, day 2, ... day 30.

[0113] Among them, multiple historical periods can also be composed of multiple cycles. For example, multiple historical periods can be the first historical period, the second historical period, and the third historical period. The first historical period is any period in the first growth cycle of the plant, the second historical period is any period in the second growth cycle of the plant, and the third historical period is any period in the third growth cycle of the plant.

[0114] In some embodiments, training samples for a machine learning model are constructed based on the growth status of plants in a greenhouse over multiple historical periods, the internal climate status of the greenhouse over multiple historical periods, and the external weather status of the greenhouse over multiple historical periods. This includes: performing the following processing for any historical period among the multiple historical periods: obtaining the growth status of the next historical period, the internal climate status of the next historical period, and the external weather status of the next historical period; combining the growth status of plants in the greenhouse over the historical period, the internal climate status of the greenhouse over the historical period, and the external weather status of the historical period into first state information for the historical period; combining the growth status of the next historical period, the internal climate status of the next historical period, and the external weather status of the next historical period into second state information for the next historical period; constructing training samples for the historical period based on the first state information of the historical period, environmental control information used to control the greenhouse during the historical period, and the second state information for the next historical period; and combining the training samples from multiple historical periods to obtain training samples for the machine learning model.

[0115] For example, when the growth state of a certain historical period is x1, the internal climate state is y1, and the external weather state is z1, then the state information for that historical period is s1[x1, y1, z1]. When the growth state of the next historical period is x2, the internal climate state is y2, and the external weather state is z2, then the state information for the next historical period is s2[x2, y2, z2]. When the environmental control information used to control the greenhouse in that historical period is a1, then the growth state s1 of that historical period, the environmental control information used to control the greenhouse in that historical period is a1, and the state information s2 of the next historical period are combined to form the training sample [s1, a1, s2] for that historical period. When there are N historical periods, the training samples are [s1, a1, s2], ..., [s... N-1 a N-1 s N ]、[s N a N [Termination state], where N is a natural number greater than 2, and the termination state represents the training termination condition.

[0116] In the process of dynamically adjusting and controlling the growth environment of plants in the greenhouse, training samples from historical periods can be stored in a cache space to collect training samples. When the number of training samples from historical periods in the cache space reaches a set threshold, multiple training samples from historical periods can be obtained from the cache space, and the machine learning model can be trained based on the training samples from multiple historical periods.

[0117] To save cache space, the cache space can be cleaned up periodically. For example, when the validity period of a training sample in the cache space expires, the training sample can be deleted. The importance of the training sample in the cache space can be obtained (for example, the degree of change in the growth state of the training sample in adjacent periods. The greater the degree of change, the better the plant growth, and the more important the training sample is). When the importance of the training sample is lower than the importance threshold, the training sample can be deleted from the cache space.

[0118] In some embodiments, a machine learning model is trained based on training samples of plant growth in a greenhouse to obtain a machine learning model for predicting environmental control information. This includes: constructing an objective function for the machine learning model based on training samples from any historical period in the training samples and labeled evaluation parameters from the historical period; updating the parameters of the machine learning model until the objective function converges; and using the updated parameters of the machine learning model when the objective function converges as the parameters of the machine learning model for predicting environmental control information.

[0119] For example, by using reinforcement learning algorithms, based on training samples from any historical period and labeled evaluation parameters of that historical period (such as the actual profit or yield of the plant in that historical period), the value of the objective function of the machine learning model can be determined. Then, it can be determined whether the value of the objective function of the machine learning model exceeds a preset threshold. When the value of the objective function of the machine learning model exceeds the preset threshold, the error signal of the machine learning model is determined based on the objective function of the machine learning model. The error information is then backpropagated in the machine learning model, and the model parameters of each layer are updated during the propagation process.

[0120] Here, we explain backpropagation. Training sample data is input into the input layer of the neural network model, passes through the hidden layers, and finally reaches the output layer to output the result. This is the forward propagation process of the neural network model. Since there is an error between the output result and the actual result, the error between the output result and the actual value is calculated and propagated back from the output layer to the hidden layers until it reaches the input layer. During the backpropagation process, the values ​​of the model parameters are adjusted according to the error. This process is iterated until convergence. The machine learning model in this context belongs to the category of neural network models.

[0121] In some embodiments, before constructing the objective function of the machine learning model, the method further includes: calling the machine learning model to perform prediction processing based on training samples from any historical period in the training samples to obtain predicted environmental control information for controlling the greenhouse in the historical period; obtaining prediction evaluation parameters for the historical period based on the predicted environmental control information; and constructing the objective function of the machine learning model based on the training samples, prediction evaluation parameters, and labeled evaluation parameters for the historical period.

[0122] For example, during the training phase, a machine learning model is invoked for prediction processing to obtain predictive environmental control information for controlling the greenhouse in historical periods. Based on this information, predictive evaluation parameters for historical periods are derived. For instance, based on the predictive environmental control information, the achievable growth states of plants in the greenhouse are predicted, and the expected growth information resulting from these states is used as the prediction evaluation parameter. Similarly, the resource information required for environmental control is predicted, and the difference between the expected growth information and the resource information is used as the annotation evaluation parameter. Finally, based on the training samples, the predictive evaluation parameters, and the annotation evaluation parameters from historical periods, the objective function of the machine learning model is constructed.

[0123] As an example, the first mapping network in the machine learning model maps the growth status of historical periods to obtain the expected growth status that meets the planting target in the next historical period. Based on the mapping relationship between the greenhouse environmental control information and external weather conditions and the internal climate state of the greenhouse, which are included in the fusion network in the machine learning model, the internal climate state of the next historical period is mapped into the predicted environmental control information used to control the greenhouse in the next historical period.

[0124] For example, based on training samples from any historical period, prediction evaluation parameters from the historical period, and labeled evaluation parameters from the historical period (such as the actual profit or yield of the plant in that historical period), after determining the value of the objective function (e.g., cross-entropy loss function) of the machine learning model, it can be determined whether the value of the objective function of the machine learning model exceeds a preset threshold. When the value of the objective function of the machine learning model exceeds the preset threshold, the error signal of the machine learning model is determined based on the objective function of the machine learning model, the error information is backpropagated in the machine learning model, and the model parameters of each layer are updated during the propagation process.

[0125] Following the example above, before constructing the objective function of the machine learning model, the process includes: obtaining the growth state of the next historical period from the training samples of the previous historical period; calling the plant simulator model based on the growth state of the next historical period to determine the expected growth information brought by the plant growth in the previous historical period; and using the expected growth information as the annotation evaluation parameter for the previous historical period. For example, the actual yield of the plant in that historical period can be used as the annotation evaluation parameter for the previous historical period.

[0126] Following the example above, before constructing the objective function of the machine learning model, the process includes: obtaining the growth status of the next historical period from the training samples of the previous historical period; invoking the plant simulator model based on the growth status of the next historical period to determine the expected growth information brought by the plant growth in the previous historical period; obtaining the environmental control information used to control the greenhouse in the previous historical period; invoking the plant simulator model based on the environmental control information used to control the greenhouse in the previous historical period to determine the resource information required for the environmental control information; and using the growth expectation information as the annotation evaluation parameter for the previous historical period, including: using the difference between the growth expectation information and the resource information as the annotation evaluation parameter for the previous historical period. For example, the actual profit brought by the plant in that historical period can be used as the annotation evaluation parameter for the previous historical period.

[0127] The following will describe an exemplary application of the embodiments of this application in a real-world greenhouse crop cultivation scenario.

[0128] With population growth and urbanization, precision agriculture has become increasingly important. Improving agricultural productivity, reducing resource consumption, meeting the needs of a larger population, and enabling people to live healthier lives through more intelligent greenhouse systems, using less manpower, is a new direction for agriculture that countries are vigorously developing. Many intelligent greenhouses have already emerged. These greenhouses use numerous sensors to monitor the environment inside and outside the greenhouse and achieve intelligent control of the crop cultivation process through heaters, lighting, drip irrigation, and other methods.

[0129] This application provides a precise and comprehensive greenhouse cultivation model capable of accurately modeling external weather and the internal greenhouse environment, as well as the complete crop growth cycle. It can even obtain minute-level greenhouse and crop status data. With the complete crop growth cycle in hand, reinforcement learning algorithms can be used to train cultivation strategies. Unlike supervised learning, reinforcement learning emphasizes environment-based actions to maximize expected benefits, seeking a balance between exploring the environment and learning from past experiences. This makes it highly suitable for planning control strategies (i.e., cultivation strategies). During the exploration process, reinforcement learning may also surpass human limitations, discovering previously undiscovered cultivation patterns to achieve greater profits.

[0130] This application provides a method and platform (system) for greenhouse crop cultivation strategies, applicable to smart agriculture, precision greenhouse control, and other fields. Users utilize this platform to establish a specific simulated greenhouse cultivation environment, where they can train and test different cultivation strategies, compare strategies, and optimize parameters. Once a mature and high-performing cultivation strategy is obtained, it can be deployed in a real agricultural greenhouse. The cultivation strategy can be based on external weather conditions, internal greenhouse conditions (internal climate conditions), and the crop's own growth status to implement precise control over multiple dimensions such as heating or cooling, CO2 release, irrigation, and lighting, thereby achieving better cultivation benefits.

[0131] like Figure 8As shown in the embodiment of this application, the platform consists of four modules: an environment module, a strategy module, a data collection module, and an experiment module. The environment module is used to create a reinforcement learning environment that simulates the real crop planting process. The strategy module provides components for implementing reinforcement learning methods and constructing reinforcement learning strategies, allowing users to create planting strategies based on this module. The data collection module controls the interaction between the strategy module and the environment module and collects data during the interaction process. The experiment module is the top-level encapsulation of the entire platform, divided into training experiments and testing experiments, and specifies the entire training and testing process. For example, during training, it defines the number of interaction steps between the strategy module and the environment module before the planting strategy self-updates, the termination condition for the end of planting strategy training, and the number of interaction steps during the testing process.

[0132] The following details the interaction flow between the environment module, strategy module, data collection module, and experiment module:

[0133] First, an environment module is used to generate a simulated agricultural planting environment. Then, a reinforcement learning agricultural planting strategy is built based on the strategy module. Next, the data collection module controls the interaction between the environment module and the strategy module, providing the planting strategy with the current state of the environment (including external weather conditions, internal greenhouse climate conditions, crop growth conditions, etc.). The planting strategy calculates an action value (i.e., environmental control information, including control measures such as heating, ventilation, irrigation, and lighting) based on this state. The simulated environment receives this action value and executes it. Afterward, the current state of the environment will transition to the next state and return a reward value (i.e., evaluation parameters, such as crop yield in the next state, profit brought by the crop in the next state, etc.) to represent the single-step effect of the current action. When the transitioned next state is the environment termination state, the current environment will be reset to the initial state. When the transitioned next state is not the environment termination state, the transitioned next state of the current environment is again used as input to the strategy module, and the above process continues, repeating in this way.

[0134] The data collection module continuously collects records such as (state value, action value, reward value, environment termination flag, and next state value) during the interaction between the strategy module and the environment module, and stores them in the data cache pool (cache space). All the records in each complete planting cycle constitute a planting trajectory. When the number of records in the data cache pool reaches a certain set value, the interaction between the strategy module and the environment module is paused, and the data in the data cache pool is used to train the planting strategy.

[0135] The training process of reinforcement learning will be further explained below. To better illustrate the training process of the planting strategy, the relevant definitions are symbolically represented as shown in Table 1:

[0136] Table 1

[0137]

[0138] The policy training process employs the standard reinforcement learning policy gradient algorithm. The goal of reinforcement learning is to obtain a planting policy that selects an optimal action based on the current state to maximize long-term value. The policy gradient algorithm directly models and optimizes this planting policy, which can be represented as... Parameterized functions This indicates that in state s, the planting strategy will provide the corresponding action value 'a'. The objective function to be maximized is to maximize the value of the action value 'a' obtained by executing the planting strategy. The objective function for the cumulative reward obtained is shown in formula (1):

[0139] (1)

[0140] in, Indicates planting strategy The resulting stationary distribution of the Markov chain (i.e., the probability of each state appearing in the state space).

[0141] To solve this maximization problem, the gradient ascent algorithm can be used to adjust the parameters. gradient By changing the given direction, the optimal path can be found. Therefore, the corresponding planting strategy The cumulative reward is maximized. After calculation and simplification, the general form of the objective function gradient can be obtained, as shown in formula (2):

[0142] (2)

[0143] The steps of the gradient policy algorithm are as follows:

[0144] Step 1: Randomly initialize planting strategy parameters

[0145] Step 2: If the strategy has not yet converged, perform the following steps:

[0146] Step a): A complete planting trajectory generated using the current planting strategy.

[0147] Step b): For each time step t, where 1 ≤ t ≤ T, perform the following steps:

[0148] Step 1: Estimate cumulative return

[0149] Step 2: Update parameters

[0150] The embodiments of this application are not limited to the training process and method described above, but may also be other training methods based on records (state values, action values, reward values, environment termination flags, and next state values) in the data cache pool.

[0151] The following details the implementation of each specific module in the platform:

[0152] (1) Environment Module

[0153] The environment module provides a simulated agricultural environment, with a greenhouse crop cultivation simulator at its core. This simulator is built on a real agricultural greenhouse, indistinguishable from a real crop greenhouse, while crop growth status is calculated using planting strategies. The simulator's operable components include 44 dimensions (action space), such as temperature settings, CO2 concentration settings, fluorescent light on / off times, and irrigation times, allowing for comprehensive and detailed control of the greenhouse. Users can also set external weather conditions. The simulator's observation space includes 38 dimensions (state space), such as external greenhouse temperature and humidity, internal greenhouse temperature, internal CO2 concentration, crop leaf area index, fruit dry weight, wet weight, planting costs, and crop profits, enabling precise descriptions of external weather, internal climate, crop condition, and economic benefits.

[0154] In summary, the simulator can provide a comprehensive description of the entire crop cultivation cycle, and its simulation process is as follows: Figure 9 As shown, the basic logic is that greenhouse control and external weather affect the greenhouse status, which in turn affects the crop growth status.

[0155] The platform in this embodiment is further encapsulated on the basis of the simulator. When using it in practice, users can select the state space and action space of the environment in the form of a configuration file, and can customize new action value functions in addition to the default action value functions. The simulated planting environment created by the environment module conforms to the OpenAI Gym interface standard.

[0156] (2) Strategy Module

[0157] The strategy module helps users build reinforcement learning planting strategies. Reinforcement learning tends to use neural networks to approximate action-value functions or state-value functions. Therefore, this strategy module provides multiple neural network components, which users can use to quickly implement reinforcement learning strategies to make fuller use of data.

[0158] In addition, the policy module also implements some reinforcement learning algorithms, such as Deep Deterministic Policy Gradient (DDPG), Advantage ActorCritic (A2C), and Twin Delayed Deep Deterministic Policy Gradient (TD3), which can be directly used for training and testing, serving as a baseline for comparative experiments.

[0159] (3) Data collection module

[0160] like Figure 10 As shown, the data collection module provides data collection and storage functions, including a data collector and a data cache (including a cache pool). The data collector defines the interaction process between the policy module and the environment module, while the data cache stores records such as state values, action values, reward values, and environment termination flags during the interaction. The data collector and data cache perform additional data processing; for example, the data collection module also provides parallel collection functionality to accelerate the collection process.

[0161] (4) Experimental Module

[0162] The experiment module is the top-level encapsulation of the entire platform, controlling the interaction logic between various modules. The experiment module includes a training section and a testing section. The training section specifies the interaction logic between the data collector and the policy module, such as how many steps to collect before updating the planting policy, and standardizes various parameters during training, such as the maximum capacity of the cache pool and the convergence threshold. The testing section controls the interaction between the testing policy module and the testing environment module, and records selected observations during the interaction.

[0163] In summary, the method and platform for greenhouse crop cultivation strategies provided in this application have the following beneficial effects:

[0164] 1) The platform provides a sophisticated greenhouse agricultural planting simulation environment and related reinforcement learning tools, which can help users quickly develop and test reinforcement learning agricultural planting strategies. The selected planting strategies can be deployed in real greenhouses to improve the planting efficiency of real greenhouses.

[0165] 2) This method has a wider strategy search space and requires almost no actual planting data or agricultural expert experience. It can learn purely through algorithms, making up for the disadvantages of agricultural planting strategy development in related technologies, reducing costs and improving development efficiency.

[0166] 3) The core of the agricultural planting environment provided by the platform is a greenhouse crop planting simulator, which provides a relatively comprehensive and accurate simulation of the real greenhouse environment and crop growth process.

[0167] This concludes the description of the artificial intelligence-based greenhouse control method provided in this application, using the exemplary application and implementation of the server provided in the embodiments of this application. This application also provides a greenhouse control device. In practical applications, the functional modules in the greenhouse control device can be collaboratively implemented using the hardware resources of electronic devices (such as terminal devices, servers, or server clusters), such as computing resources like processors, communication resources (such as those used to support various communication methods like optical cables and cellular networks), and memory. Figure 2 The greenhouse control device 555 stored in the memory 550 is shown. It can be software in the form of programs and plug-ins, such as software modules designed in programming languages ​​such as C / C++ and Java, application software designed in programming languages ​​such as C / C++ and Java, or dedicated software modules, application programming interfaces, plug-ins, cloud services, etc. in large software systems. Examples of different implementation methods are given below.

[0168] Example 1: Greenhouse control devices are mobile applications and modules.

[0169] The greenhouse control device 555 in this embodiment can provide a software module designed using programming languages ​​such as C / C++ and Java, which can be embedded into various mobile applications based on systems such as Android or iOS (stored as executable instructions in the storage medium of the mobile device and executed by the processor of the mobile device). This allows the device to directly use its own computing resources to complete related information recommendation tasks and periodically or irregularly transmit the processing results to a remote server through various network communication methods, or save them locally on the mobile device.

[0170] Example 2: The greenhouse control device is a server application and platform.

[0171] The greenhouse control device 555 in this embodiment can be provided as a dedicated software module in an application software or large software system designed using programming languages ​​such as C / C++ and Java. It runs on the server side (stored in the server-side storage medium as executable instructions and executed by the server-side processor). The server uses its own computing resources to complete the relevant information recommendation tasks.

[0172] This application embodiment can also provide an information recommendation platform (for recommendation lists) for use by individuals, groups or organizations by mounting a customized, easy-to-interact web interface or other user interfaces (UI) on a distributed, parallel computing platform composed of multiple servers.

[0173] Example 3: The greenhouse control device consists of a server-side application programming interface (API) and plugins.

[0174] The greenhouse control device 555 in this application embodiment can be provided as a server-side API or plugin for users to call, to execute the artificial intelligence-based greenhouse control method of this application embodiment, and to be embedded into various applications.

[0175] Example 4: Greenhouse control devices are mobile device client APIs and plugins.

[0176] The greenhouse control device 555 in this application embodiment can be provided as an API or plugin on a mobile device for users to call in order to execute the artificial intelligence-based greenhouse control method of this application embodiment.

[0177] Example 5: Greenhouse control devices are cloud-based open services.

[0178] The greenhouse control device 555 in this embodiment can provide a cloud service for information recommendation developed for users, allowing individuals, groups, or organizations to obtain a recommendation list.

[0179] The greenhouse control device 555 includes a series of modules, including an acquisition module 5551, a processing module 5552, an application module 5553, a training module 5554, and a storage module 5555. The following description further illustrates how the various modules in the greenhouse control device 555 provided in this embodiment cooperate to implement the greenhouse control scheme.

[0180] The acquisition module 5551 is used to acquire the growth status of plants in the greenhouse in a first period, the internal climate status of the greenhouse in the first period, and the external weather status of the greenhouse in the first period; the processing module 5552 is used to call a machine learning model based on the growth status, the internal climate status, and the external weather status in the first period to obtain environmental control information for controlling the greenhouse in a second period, wherein the second period is later than the first period; the application module 5553 is used to apply the environmental control information to the greenhouse in the second period.

[0181] In some embodiments, the processing module 5552 is further configured to perform the following processing based on the machine learning model: based on the growth state of the first period, determine the expected growth state in the second period that meets the planting target, and determine the internal climate state of the second period when the expected growth state is achieved; based on the characteristics of the external weather state of the second period and the environmental control information used to control the greenhouse in the second period acting together on the internal climate state of the second period, determine the environmental control information used to control the greenhouse in the second period.

[0182] In some embodiments, the machine learning model includes a first mapping network and a fusion network; the processing module 5552 is further configured to perform mapping processing on the growth state of the first period based on the first mapping network to obtain the expected growth state that meets the planting target in the second period; based on the mapping relationship between the environmental control information of the greenhouse and the external weather state included in the fusion network and the internal climate state of the greenhouse, the internal climate state of the second period is mapped to the environmental control information used to control the greenhouse in the second period.

[0183] In some embodiments, the machine learning model further includes a second mapping network; the processing module 5552 is further configured to perform mapping processing on the expected growth state of the second period based on the mapping relationship between the internal climate state of the greenhouse and the growth state of the plant included in the second mapping network, to obtain the internal climate state of the second period when the expected growth state of the second period is achieved; or, based on the mapping relationship between the internal climate state of the greenhouse and the change in the growth state of the plant included in the second mapping network, perform state transformation processing on the state difference between the growth state of the first period and the expected growth state of the second period, to obtain the internal climate state of the second period when the expected growth state of the second period is achieved.

[0184] In some embodiments, the fusion network includes a first convolutional layer, a second convolutional layer, a fully connected layer, and a third convolutional layer; the processing module 5552 is further configured to perform convolution processing on the external weather state of the second period based on the first convolutional layer to obtain first state information corresponding to the external weather state; perform convolution processing on the internal climate state of the second period based on the second convolutional layer to obtain second state information corresponding to the internal climate state; determine the difference between the second state information and the first state information based on the fully connected layer; and perform convolution processing on the difference based on the third convolutional layer to obtain environmental control information for controlling the greenhouse in the second period.

[0185] In some embodiments, the greenhouse control device 555 further includes: a training module 5554, configured to construct training samples for the machine learning model based on the growth status of plants in the greenhouse during multiple historical periods, the internal climate status of the greenhouse during the multiple historical periods, and the external weather status of the greenhouse during the multiple historical periods; and to train the machine learning model based on the training samples to obtain the machine learning model for predicting environmental control information.

[0186] In some embodiments, the training module 5554 is further configured to perform the following processing for any one of the plurality of historical periods: obtaining the growth state of the next historical period, the internal climate state of the next historical period, and the external weather state of the next historical period; combining the growth state of the plants in the greenhouse in the historical period, the internal climate state of the greenhouse in the historical period, and the external weather state of the historical period into first state information of the historical period; combining the growth state of the next historical period, the internal climate state of the next historical period, and the external weather state of the next historical period into second state information of the next historical period; constructing training samples for the historical period based on the first state information of the historical period, the environmental control information used to control the greenhouse in the historical period, and the second state information of the next historical period; and combining the training samples of the plurality of historical periods to obtain training samples for the machine learning model.

[0187] In some embodiments, the greenhouse control device 555 further includes: a storage module 5555, configured to store training samples from the historical period into a cache space; the training module 5554 is further configured to, when the number of training samples from the historical period in the cache space reaches a set threshold, obtain multiple training samples from the historical period from the cache space, and train the machine learning model based on the multiple training samples from the historical period.

[0188] In some embodiments, the training module 5554 is further configured to construct the objective function of the machine learning model based on training samples from any historical period in the training samples and the labeled evaluation parameters of the historical period; update the parameters of the machine learning model until the objective function converges; and use the updated parameters of the machine learning model when the objective function converges as the parameters of the machine learning model for predicting environmental control information.

[0189] In some embodiments, the processing module 5552 is further configured to invoke the machine learning model to perform prediction processing based on training samples from any historical period in the training samples, to obtain predicted environmental control information for controlling the greenhouse in the historical period; and to obtain prediction evaluation parameters for the historical period based on the predicted environmental control information; the training module 5554 is further configured to construct the objective function of the machine learning model based on the training samples of the historical period, the prediction evaluation parameters of the historical period, and the labeled evaluation parameters of the historical period.

[0190] In some embodiments, the application module 5553 is further configured to obtain the growth status of the next historical period in the training samples of the historical period; call the plant simulator model based on the growth status of the next historical period to determine the growth expectation information brought about by the plant growth in the historical period, and use the growth expectation information as the annotation evaluation parameter of the historical period.

[0191] In some embodiments, the application module 5553 is further configured to acquire environmental control information used to control the greenhouse during the historical period; invoke the plant simulator model based on the environmental control information used to control the greenhouse during the historical period to determine the resource information required by the environmental control information; and use the difference between the growth expectation information and the resource information as a labeling evaluation parameter for the historical period.

[0192] This application provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the artificial intelligence-based greenhouse control method described above in this application.

[0193] This application provides a computer-readable storage medium storing executable instructions. When these executable instructions are executed by a processor, they cause the processor to execute the artificial intelligence-based greenhouse control method provided in this application. For example... Figures 3-5 The greenhouse control method based on artificial intelligence is shown.

[0194] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of devices including one or any combination of the above-mentioned memories.

[0195] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0196] As an example, executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborating files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0197] As an example, executable instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0198] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. An artificial intelligence-based greenhouse control method, characterized by, The method comprises: acquiring a growth state of a plant in a greenhouse at a first period, an internal climate state of the greenhouse at the first period, and an external weather state of the greenhouse at the first period; mapping the growth state at the first period based on a first mapping network in a machine learning model to obtain an expected growth state that meets a planting target in a second period, the second period being later than the first period; determining an internal climate state of the second period when the expected growth state is achieved; mapping the internal climate state of the second period to environmental control information for controlling the greenhouse in the second period based on a mapping relationship between the environmental control information of the greenhouse and the internal climate state of the greenhouse and the external weather state of the greenhouse included in a fusion network in the machine learning model; applying the environmental control information to the greenhouse in the second period.

2. The method of claim 1, wherein: the machine learning model further comprises a second mapping network; the determination of the internal climate state of the second period when the expected growth state is achieved comprises: mapping the expected growth state of the second period based on a mapping relationship between the internal climate state of the greenhouse and the growth state of the plant included in the second mapping network to obtain the internal climate state of the second period when the expected growth state of the second period is achieved; or performing state conversion processing on a state difference between the growth state at the first period and the expected growth state of the second period based on a mapping relationship between the internal climate state of the greenhouse and a growth state change of the plant included in the second mapping network to obtain the internal climate state of the second period when the expected growth state of the second period is achieved.

3. The method of claim 1, wherein: the fusion network comprises a first convolutional layer, a second convolutional layer, a fully connected layer, and a third convolutional layer; the mapping of the internal climate state of the second period to the environmental control information for controlling the greenhouse in the second period comprises: performing convolution processing on the external weather state of the second period based on the first convolutional layer to obtain first state information corresponding to the external weather state; performing convolution processing on the internal climate state of the second period based on the second convolutional layer to obtain second state information corresponding to the internal climate state; determining a difference between the second state information and the first state information based on the fully connected layer; performing convolution processing on the difference based on the third convolutional layer to obtain the environmental control information for controlling the greenhouse in the second period.

4. The method of claim 1, wherein, The method further comprises: constructing a training sample of the machine learning model based on growth states of plants in the greenhouse at a plurality of historical periods, internal climate states of the greenhouse at the plurality of historical periods, and external weather states of the greenhouse at the plurality of historical periods; training the machine learning model based on the training sample to obtain the machine learning model for environmental control information prediction.

5. The method of claim 4, wherein, The training sample of the machine learning model is constructed based on the growth state of the plants in the greenhouse at a plurality of historical periods, the internal climate state of the greenhouse at the plurality of historical periods, and the external weather state of the greenhouse at the plurality of historical periods, and the method comprises the following steps: The following processing is performed for any one of the plurality of historical periods: Obtaining the growth state of the next historical period of the historical period, the internal climate state of the next historical period of the historical period, and the external weather state of the next historical period of the historical period; Combining the growth state of the plants in the greenhouse at the historical period, the internal climate state of the greenhouse at the historical period, and the external weather state of the historical period into first state information of the historical period; Combining the growth state of the next historical period of the historical period, the internal climate state of the next historical period of the historical period, and the external weather state of the next historical period of the historical period into second state information of the next historical period; Constructing a training sample of the historical period based on the first state information of the historical period, the environmental control information for controlling the greenhouse in the historical period, and the second state information of the next historical period; Combining the training samples of a plurality of historical periods to obtain the training sample of the machine learning model.

6. The method of claim 5, wherein the method further comprises: storing the training sample of the historical period to a cache space; training the machine learning model based on the training sample comprises: when the number of training samples of the historical period in the cache space reaches a set threshold, obtaining a plurality of training samples of the historical period from the cache space, and training the machine learning model based on a plurality of training samples of the historical period. training the machine learning model based on the training sample to obtain the machine learning model for environmental control information prediction comprises:

7. The method of claim 5, wherein, constructing an objective function of the machine learning model based on the training sample of any one of the historical periods and the labeled evaluation parameter of the historical period; updating the parameters of the machine learning model until the objective function converges, and using the updated parameters of the machine learning model when the objective function converges as the parameters of the machine learning model for environmental control information prediction.

8. The method of claim 7, wherein before constructing the objective function of the machine learning model, the method further comprises: calling the machine learning model to perform prediction processing based on the training sample of any one of the historical periods, to obtain predicted environmental control information for controlling the greenhouse in the historical period; obtaining a predicted evaluation parameter of the historical period based on the predicted environmental control information; constructing the objective function of the machine learning model comprises: constructing the objective function of the machine learning model based on the training sample of the historical period, the predicted evaluation parameter of the historical period, and the labeled evaluation parameter of the historical period. ​ ​ 9. The method according to claim 7 or 8, characterized in that, The method further includes, before constructing the objective function of the machine learning model: obtaining a growth state of a next historical period in the training samples of the historical period; calling a plant simulator model based on the growth state of the next historical period to determine growth expectation information brought by the plant growth in the historical period, and taking the growth expectation information as a labeled evaluation parameter of the historical period.

10. The method of claim 9, wherein, The method further includes, before taking the growth expectation information as the labeled evaluation parameter of the historical period: obtaining environmental control information for controlling the greenhouse in the historical period; calling the plant simulator model based on the environmental control information for controlling the greenhouse in the historical period to determine resource information consumed by the environmental control information; The method further includes, taking the growth expectation information as the labeled evaluation parameter of the historical period: taking a difference between the growth expectation information and the resource information as the labeled evaluation parameter of the historical period.

11. A greenhouse control device, characterized by The apparatus includes: an obtaining module, configured to obtain a growth state of a plant in a first period in a greenhouse, an internal climate state of the greenhouse in the first period, and an external weather state of the greenhouse in the first period; a processing module, configured to perform mapping processing on the growth state of the first period based on a first mapping network in a machine learning model, to obtain an expected growth state meeting a planting target in a second period later than the first period, to determine an internal climate state of the second period when the expected growth state is achieved, and to map the internal climate state of the second period to environmental control information for controlling the greenhouse in the second period based on a mapping relationship between the environmental control information of the greenhouse and the internal climate state of the greenhouse and the external weather state of the greenhouse included in a fusion network in the machine learning model; an applying module, configured to apply the environmental control information to the greenhouse in the second period.

12. The apparatus of claim 11, wherein, The machine learning model further includes a second mapping network. The processing module is further configured to: perform mapping processing on the expected growth state of the second period based on a mapping relationship between the internal climate state of the greenhouse and the growth state of the plant included in the second mapping network, to obtain the internal climate state of the second period when the expected growth state of the second period is achieved; or perform state conversion processing on a state difference between the growth state of the first period and the expected growth state of the second period based on a mapping relationship between the internal climate state of the greenhouse and a growth state change of the plant included in the second mapping network, to obtain the internal climate state of the second period when the expected growth state of the second period is achieved.

13. The apparatus of claim 11, wherein, The fusion network includes a first convolutional layer, a second convolutional layer, a fully connected layer, and a third convolutional layer. The processing module is further configured to: perform convolution processing on the external weather state of the second period based on the first convolutional layer, to obtain first state information corresponding to the external weather state; convolve, based on the second convolutional layer, the internal climate state of the second period to obtain second state information corresponding to the internal climate state; determine, based on the fully connected layer, a difference between the second state information and the first state information; convolve, based on the third convolutional layer, the difference to obtain environmental control information for controlling the greenhouse in the second period.

14. The apparatus of claim 11, wherein, The apparatus further includes a training module configured to: construct, based on the growth state of the plants in the greenhouse in a plurality of historical periods, the internal climate state of the greenhouse in the plurality of historical periods, and the external weather state of the greenhouse in the plurality of historical periods, a training sample of the machine learning model; train, based on the training sample, the machine learning model to obtain the machine learning model for predicting environmental control information.

15. The apparatus of claim 14, wherein, The training module is further configured to: for any historical period in the plurality of historical periods, perform the following processing: obtain the growth state of a next historical period of the historical period, the internal climate state of the next historical period of the historical period, and the external weather state of the next historical period of the historical period; combine the growth state of the plants in the greenhouse in the historical period, the internal climate state of the greenhouse in the historical period, and the external weather state of the historical period into first state information of the historical period; combine the growth state of the next historical period of the historical period, the internal climate state of the next historical period of the historical period, and the external weather state of the next historical period of the historical period into second state information of the next historical period; construct, based on the first state information of the historical period, the environmental control information for controlling the greenhouse in the historical period, and the second state information of the next historical period, a training sample of the historical period; combine the training samples of the plurality of historical periods to obtain the training sample of the machine learning model.

16. The apparatus of claim 15, wherein: the apparatus further includes a storage module configured to store the training sample of the historical period to a cache space; the training module is further configured to, when the number of the training samples of the historical period in the cache space reaches a set threshold, obtain the training samples of the plurality of historical periods from the cache space and train the machine learning model based on the training samples of the plurality of historical periods.

17. The apparatus of claim 15, wherein, The training module is further configured to: construct, based on the training sample of any historical period in the training sample and a labeled evaluation parameter of the historical period, an objective function of the machine learning model; update parameters of the machine learning model until the objective function converges, and use the updated parameters of the machine learning model when the objective function converges as the parameters of the machine learning model for predicting environmental control information.

18. An electronic device, comprising: The electronic device includes: a memory configured to store executable instructions; a processor configured to execute the executable instructions stored in the memory to implement the artificial intelligence-based greenhouse control method of any one of claims 1 to 10.

19. A computer-readable storage medium, characterized in that, A computer program product comprising computer instructions for implementing the artificial intelligence based greenhouse control method of any one of claims 1 to 10 when executed by a processor.

20. A computer program product, characterised in that, A computer program product comprising computer instructions for implementing the artificial intelligence based greenhouse control method of any one of claims 1 to 10 when executed by a processor.

Images