Greenhouse environment regulation and control method and device, electronic equipment and storage medium
By combining environmental prediction models and large language models, the target control commands for the greenhouse environment are determined, which solves the problems of insufficient flexibility and decentralized decision-making in traditional greenhouse environmental control systems, and achieves more rational and efficient greenhouse environmental regulation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-30
- Publication Date
- 2026-04-21
AI Technical Summary
Traditional greenhouse environmental control systems lack flexibility and cannot effectively handle complex coupling relationships, leading to control conflicts or poor performance. Furthermore, the independent operation of prediction models and rules results in fragmented decision-making, affecting overall optimality.
By combining environmental prediction models and large language models, and based on multiple candidate equipment action schemes, the most suitable target control command is determined. By combining action scheme prediction models and large language models, the impact of candidate equipment action schemes on the greenhouse environment is considered, so as to achieve more reasonable regulation.
It enables precise control of the greenhouse environment, avoids control conflicts, improves the overall optimality and adaptability of control, and enhances the ability to handle complex coupling relationships.
Smart Images
Figure CN121900546A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of greenhouse environment control technology, specifically to a greenhouse environment control method, a greenhouse environment control device, electronic equipment, a storage medium, and a computer program product. Background Technology
[0002] In recent years, with the development of facility agriculture and smart agriculture, greenhouse environmental control systems have gradually transitioned from manual, experience-based regulation to automated and intelligent control. The internal environmental parameters of a greenhouse (temperature, humidity, light intensity, carbon dioxide concentration, etc.) are not only affected by external climatic conditions but also exhibit complex coupling relationships. For example, ventilation and cooling can lead to a decrease in carbon dioxide concentration, supplemental lighting can increase the temperature inside the greenhouse, and humidification can affect humidity and transpiration rates. Therefore, accurate control of the greenhouse environment is becoming increasingly important.
[0003] Traditional greenhouse environmental control typically relies on thresholds or rules, which lacks flexibility and cannot handle the complex coupling relationships within the greenhouse environment. When multiple environmental parameters fluctuate simultaneously, control conflicts or ineffective results can easily occur. While predictive models and rules-based greenhouse environmental control systems exist in traditional technologies, these models and rules usually operate independently, failing to integrate solutions from different sources. This leads to fragmented decision-making, affecting the overall optimality of the final control scheme. Summary of the Invention
[0004] The present invention was proposed in view of the above-mentioned problems.
[0005] According to a first aspect of the present invention, a method for controlling the environment of a greenhouse is provided. The method includes: determining an environmental prediction result for the target greenhouse in a future preset time period based on environmental parameters and equipment operating status of the target greenhouse during a preset time period prior to the current moment, wherein the equipment operating status of the target greenhouse affects the environmental parameters of the target greenhouse; inputting the environmental prediction result and preset knowledge rules into an action plan prediction model to obtain multiple candidate equipment action plans for the target greenhouse; using a large language model, determining a target control command corresponding to one of the multiple candidate equipment action plans; and performing control operations on the equipment of the target greenhouse according to the target control command.
[0006] For example, the large language model includes a first large language model and a second large language model. The step of using the large language model to determine the target control command corresponding to one of the multiple candidate device action schemes includes: inputting the multiple candidate device action schemes into the first large language model respectively to output their respective initial control commands; and using the second large language model to determine the target control command from the initial control commands.
[0007] For example, determining the target control instruction from the initial control instructions using the second large language model includes: inputting a preset mutual exclusion rule and the initial control instruction into the second large language model to output a recognition result, wherein the recognition result indicates whether the initial control instruction includes a mutual exclusion instruction that conforms to the preset mutual exclusion rule; and selecting a target control instruction from the initial control instructions based on the recognition result, wherein the recognition result of the target control instruction indicates that the target control instruction does not include a mutual exclusion instruction.
[0008] Exemplarily, the method further includes:
[0009] When the identification results of all initial control commands indicate that the initial control command includes a mutual exclusion command, a reference control command corresponding to the reference prediction result is retrieved from the historical record as the current target control command. The historical record includes corresponding stored reference prediction results and reference control commands, and the reference prediction result is the environmental prediction result with the smallest deviation from the environmental prediction result in the historical record.
[0010] For example, the method further includes: counting the number of times the step of retrieving the reference control instruction corresponding to the reference prediction result from the historical record is performed within a preset time window; when the number of operations is greater than or equal to a preset number threshold, using a proximal policy optimization algorithm, based on the environmental parameters and the environmental prediction result, to perform reinforcement learning training on the action plan prediction model.
[0011] For example, the step of inputting the multiple sets of candidate device action schemes into the first large language model to output their respective initial control commands includes: inputting the multiple sets of candidate device action schemes into the first large language model to output their respective initial control commands and their respective natural language descriptions.
[0012] For example, the method further includes: determining the target execution result of each of the multiple candidate device action schemes based on the environmental parameters; and determining a target control instruction corresponding to one of the multiple candidate device action schemes using a large language model, which includes: determining the initial control instruction corresponding to each of the multiple candidate device action schemes using the large language model; calculating a comprehensive index corresponding to the initial control instruction based on the target difference between the target execution result of the candidate device action scheme corresponding to each initial control instruction and the suitable environmental parameters of the target plant in the target greenhouse; and selecting the target control instruction from the initial control instructions based on the comprehensive index.
[0013] For example, the calculation of the comprehensive index corresponding to the initial control command is also based on the equipment energy consumption and / or equipment start-stop smoothness of the candidate equipment action scheme corresponding to the initial control command.
[0014] For example, the method further includes: for the target execution result of the candidate device action scheme corresponding to each initial control command, determining a weight coefficient corresponding to each target difference based on the growth stage of the target plant, wherein the target differences are the differences between the target execution result of the same type and the suitable environmental parameters of the growth stage of the target plant; calculating a weighted difference value for each target difference based on the weight coefficient and the target difference value; and determining the target difference between the target execution result of the candidate device action scheme corresponding to the initial control command and the suitable environmental parameters of the growth stage of the target plant based on the weighted difference value of each target difference value.
[0015] For example, determining the environmental prediction result of the target greenhouse in the future preset period based on the environmental parameters and equipment operating status of the target greenhouse in the preset period before the current time includes: inputting the environmental parameters and the equipment operating status into the environmental prediction model to output the environmental prediction result;
[0016] The method further includes: determining the target execution result of each of the multiple candidate device action schemes based on the environmental parameters; determining the deviation between the actual environmental parameters and the theoretical execution result after executing the control operation, wherein the theoretical execution result is the target execution result of the candidate device action scheme corresponding to the target control command; and adjusting the parameters of the environmental prediction model based on a preset training strategy when the deviation is greater than a preset deviation threshold.
[0017] For example, the method further includes updating the preset threshold based on the time period in which the current moment is located and the distribution of the deviation over the time period.
[0018] According to a second aspect of the present invention, a greenhouse environment control device is also provided, comprising:
[0019] The first prediction module is used to determine the environmental prediction result of the target greenhouse in the future preset period based on the environmental parameters and equipment operating status of the target greenhouse in the preset period before the current time, wherein the equipment operating status of the target greenhouse affects the environmental parameters of the target greenhouse;
[0020] The second prediction module is used to input the environmental prediction results and preset knowledge rules into the action plan prediction model to obtain multiple candidate equipment action plans for the target greenhouse.
[0021] The third prediction module is used to determine a target control command corresponding to one of the multiple candidate device action schemes based on the large language model.
[0022] The control module is used to perform control operations on the equipment of the target greenhouse according to the target control command.
[0023] According to a third aspect of the present invention, an electronic device is also provided, comprising: a processor and a memory, wherein the memory stores computer program instructions, which, when executed by the processor, are used to perform the above-described greenhouse environment control method.
[0024] According to a fourth aspect of the present invention, a storage medium is also provided, on which program instructions are stored, which, when executed, are used to perform the above-described greenhouse environment control method.
[0025] According to a fifth aspect of the present invention, a computer program product is also provided, comprising computer program instructions, which, when executed, are used to perform the above-described greenhouse environment control method.
[0026] In the above technical solution, based on the environmental parameters and equipment operating status of the target greenhouse during a preset time period prior to the current moment, the environmental prediction results for the target greenhouse during a future preset time period are determined. The equipment operating status of the target greenhouse affects its environmental parameters. Then, the environmental prediction results and preset knowledge rules are input into an action plan prediction model to obtain multiple candidate equipment action plans for the target greenhouse. Next, using a large language model, a target control command corresponding to one of the multiple candidate equipment action plans is determined. Finally, control operations are performed on the equipment of the target greenhouse according to the target control command. By combining the action plan prediction model and the large language model, the impact of candidate equipment action plans on the greenhouse environment can be considered more comprehensively, resulting in the target control command most suitable for the current greenhouse environment. Therefore, the greenhouse environment can be more rationally regulated based on the target control command.
[0027] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description
[0028] The above and other objects, features, and advantages of the present invention will become more apparent from the more detailed description of the embodiments of the invention in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same parts or steps.
[0029] Figure 1 A schematic flowchart of a greenhouse environment control method according to an embodiment of the present invention is shown;
[0030] Figure 2 A schematic flowchart illustrating the determination of a target control command corresponding to one of a plurality of candidate device action schemes according to an embodiment of the present invention is shown.
[0031] Figure 3 A schematic flowchart illustrating the selection of a target control command from an initial control command according to an embodiment of the present invention is shown;
[0032] Figure 4 A schematic flowchart illustrating reinforcement training of a motion plan prediction model according to an embodiment of the present invention is shown.
[0033] Figure 5 A schematic flowchart illustrating the selection of a target control command from an initial control command based on comprehensive indicators according to an embodiment of the present invention is shown.
[0034] Figure 6 A schematic flowchart illustrating the determination of target differences according to an embodiment of the present invention is shown;
[0035] Figure 7 A schematic flowchart of a retrained environment prediction model according to an embodiment of the present invention is shown;
[0036] Figure 8 A schematic block diagram of a greenhouse environment control device according to an embodiment of the present invention is shown;
[0037] Figure 9 A schematic block diagram of an electronic device according to an embodiment of the present invention is shown. Detailed Implementation
[0038] To make the objectives, technical solutions, and advantages of the present invention more apparent, exemplary embodiments according to the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely a part of the embodiments of the present invention, and not all of the embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein. Based on the embodiments of the present invention described herein, all other embodiments obtained by those skilled in the art without inventive effort should fall within the protection scope of the present invention.
[0039] To at least partially address the aforementioned problems, a greenhouse environment control method is proposed. This method determines the environmental prediction results for the target greenhouse within a future preset time period based on the environmental parameters and equipment operating status of the target greenhouse during a preset time period prior to the current moment. The equipment operating status of the target greenhouse influences its environmental parameters. The environmental prediction results and preset knowledge rules are then input into an action plan prediction model to obtain multiple candidate equipment action plans for the target greenhouse. Next, using a large language model, a target control command corresponding to one of the multiple candidate equipment action plans is determined. Finally, control operations are performed on the equipment of the target greenhouse according to the target control command. By combining the action plan prediction model and the large language model, the impact of candidate equipment action plans on the greenhouse environment can be more comprehensively considered, resulting in the target control command most suitable for the current greenhouse environment. Therefore, the greenhouse environment can be more rationally controlled based on the target control command.
[0040] Figure 1 A schematic flowchart of a greenhouse environment control method according to an embodiment of the present invention is shown. Figure 1 As shown, the greenhouse environment control method may include steps S110 to S140.
[0041] In step S110, based on the environmental parameters and equipment operating status of the target greenhouse during a preset time period prior to the current time, the environmental prediction result of the target greenhouse during a future preset time period is determined, wherein the equipment operating status of the target greenhouse affects the environmental parameters of the target greenhouse.
[0042] The environmental parameters of the target greenhouse can include parameters such as temperature, humidity, carbon dioxide concentration, and light intensity. The environmental prediction result for the target greenhouse within a preset future time period is the predicted environmental information based on the equipment operating status within that preset time period. The types of environmental parameters in the environmental prediction result and the environmental parameters of the target greenhouse can be the same. The preset time period can be the same as or different from the duration of the future preset time period. For example, the environmental parameters and equipment operating status of the target greenhouse within 3 minutes before the current moment can be used to predict the environmental prediction result within 10 minutes before and after the current moment.
[0043] For example, the types of environmental parameters for the target greenhouse in the environmental prediction results and the preset time period prior to the current time can be the same. For instance, if the environmental parameters for the target greenhouse in the preset time period prior to the current time include temperature, the environmental prediction results can also include temperature.
[0044] The equipment in the target greenhouse may include side fans, circulating fans, evaporative cooling pads, roof windows, internal insulation, shading awnings, and external insulation. Correspondingly, the operating status of the equipment in the target greenhouse may include its on / off status, opening degree, and energy consumption. For example, for fans, the operating status of the target greenhouse equipment may include the fan's on / off status, speed, and operating power.
[0045] For example, based on the environmental parameters of the target greenhouse during a preset time period prior to the current moment, the changes in the target greenhouse's environmental parameters over time can be determined. Then, based on these changes and the time of the next preset time period, the environmental prediction results for the target greenhouse during that future preset time period can be determined.
[0046] For example, the operating status of the target greenhouse's equipment can be used as a latent or exogenous variable to establish an autoregressive model of the target greenhouse's environmental parameters over a preset period prior to the current moment. Based on this autoregressive model, environmental predictions for the target greenhouse over a future preset period can be calculated.
[0047] For example, the operating status of the target greenhouse's equipment and its environmental parameters within a preset time period prior to the current moment can be input into a time-series prediction model to output an environmental prediction result for the target greenhouse within a future preset time period. For instance, a time-series prediction model based on a Long Short-Term Memory (LSTM) network can be used to predict the environmental prediction result for the target greenhouse within a future preset time period. Training data on the operating status of the target greenhouse's equipment and its environmental parameters can be obtained in advance, and then the time-series prediction model can be trained using this training data to obtain the trained time-series prediction model.
[0048] In step S120, the environmental prediction results and preset knowledge rules are input into the action plan prediction model to obtain multiple candidate equipment action plans for the target greenhouse.
[0049] The preset knowledge rules can include pre-defined knowledge related to the control actions of the equipment in the target greenhouse. For example, the preset knowledge rules can include the correspondence between environmental prediction results and the control actions of the equipment in the target greenhouse. Another example is the correspondence between target differences and the control actions of the equipment in the target greenhouse, where the target differences could be the difference between the environmental prediction results and the suitable environmental parameter ranges for the target plants within the greenhouse.
[0050] The action plan prediction model can analyze the environmental prediction results based on preset knowledge rules to determine the control actions that may be needed, thus obtaining multiple candidate equipment action plans for the target greenhouse. Each candidate equipment action plan can include at least one equipment action, such as turning the fan on by 30%, opening the top window, or turning off the humidifier. For example, in addition to the equipment action, each candidate equipment action plan may also include the execution time of the equipment action; for instance, a candidate equipment action plan could include turning the fan on by 30% after 3 minutes.
[0051] The time period targeted by the candidate device action plan can be a preset future time period. For example, each candidate device action plan can include device actions within 10 minutes after the current time.
[0052] For example, the time period targeted by the candidate device action plan can be a future preset time period. For instance, if the future preset time period is 10 minutes after the current time, the candidate device action plan can include device actions within 10 minutes after the current time.
[0053] For example, the candidate device action plan may include either a specifically described device action or a roughly described device action. For instance, a specifically described device action might be "turn the fan on to 80%", while a roughly described device action might be "strong ventilation". The roughly described device action corresponds to the specific device action.
[0054] For example, preset knowledge rules and environmental parameters can be used as training data. This training data is then input into an action plan prediction model to obtain predicted device action plans. The predicted device action plans can then be compared with the standard device action plans corresponding to the input training data. A loss value is calculated based on the difference between the two, and the parameters of the action plan prediction model are adjusted according to the loss value to obtain a trained plan prediction model used to predict candidate device action plans.
[0055] In step S130, using a large language model, a target control command corresponding to one of the multiple candidate device action schemes is determined based on multiple candidate device action schemes.
[0056] The large language model can be of any type. By calling the interface of the large language model, the target control command corresponding to one of multiple candidate device action schemes can be determined.
[0057] For example, multiple candidate device action schemes can be input into a large language model, so that the large language model can output a corresponding target control command from the multiple candidate device action schemes.
[0058] For example, candidate device action schemes can be input into a large language model to obtain their respective control commands, and then one of them can be determined as the target control command based on experience.
[0059] For example, candidate device action schemes can be input into a large language model to obtain their corresponding control commands. These control commands can then be input into the large language model to output one as the target control command.
[0060] In step S140, control operations are performed on the equipment of the target greenhouse according to the target control command.
[0061] After determining the target control command, the equipment targeted by the command and its corresponding control parameters can be analyzed. Then, appropriate control operations can be performed on the equipment in the target greenhouse based on the control parameters.
[0062] In the above technical solution, based on the environmental parameters and equipment operating status of the target greenhouse during a preset time period prior to the current moment, the environmental prediction results for the target greenhouse during a future preset time period are determined. The equipment operating status of the target greenhouse affects its environmental parameters. Then, the environmental prediction results and preset knowledge rules are input into an action plan prediction model to obtain multiple candidate equipment action plans for the target greenhouse. Next, using a large language model, a target control command corresponding to one of the multiple candidate equipment action plans is determined. Finally, control operations are performed on the equipment of the target greenhouse according to the target control command. By combining the action plan prediction model and the large language model, the impact of candidate equipment action plans on the greenhouse environment can be considered more comprehensively, resulting in the target control command most suitable for the current greenhouse environment. Therefore, the greenhouse environment can be more rationally regulated based on the target control command.
[0063] For example, the large language model includes the first large language model and the second large language model. Figure 2 A schematic flowchart illustrating the determination of a target control command corresponding to one of a plurality of candidate device action schemes according to an embodiment of the present invention is shown. Figure 2 As shown, step S130 may include steps S210 to S220.
[0064] In step S210, multiple candidate device action schemes are input into the first large language model to output their respective initial control commands.
[0065] For each candidate device action plan, the interface of the first language model can be called to input the candidate device action plan into the first language model, so that its corresponding initial control command in a preset format is output. For example, the preset format can be a standardized JSON control command format. The preset format can also be other types of formats, which will not be detailed here. For example, the initial control command can be used to input multiple candidate device action plans into the first language model separately, so that each outputs its corresponding initial control command and its corresponding natural language description. When outputting the initial control command, the first language model can also be used to output the natural language description corresponding to the candidate device action plan. This facilitates operators' understanding and traceability of the candidate device action plans determined by the first language model.
[0066] The initial control command is a specific control command for the equipment. Even if the equipment action in the candidate equipment action plan is described roughly or specifically, the corresponding part of the determined initial control command should be specifically described. This ensures that when used as a target command later, the equipment can be accurately controlled based on the command content. For example, if the equipment action in the candidate equipment action plan is "forced ventilation," then the corresponding initial control command could be "fan starts at 80%."
[0067] In step S220, the target control command is determined from the initial control command using the second language model.
[0068] For example, initial control commands can be input into a second language model so that the target control commands can be directly output from the second language model.
[0069] For example, initial control commands can be input into a second large language model for evaluation, outputting corresponding evaluation results. Then, based on the requirements corresponding to the evaluation results, it can be determined whether each initial control command can be used as a target control command. For instance, the evaluation result could be a comprehensive score of the initial control commands, and the initial control command with the highest comprehensive score could be selected as the target control command.
[0070] Optionally, the first and second largest language models can be large language models of the same type.
[0071] Optionally, the first and second large language models can be different types of large language models to avoid homogenization of recognition results.
[0072] In the above technical solution, multiple candidate device action schemes are input into a first large language model to output their respective initial control commands. Then, a second large language model is used to determine the target control command from the initial control commands. Compared to determining the target control command based on experience, using both the first and second large language models allows for the determination of a more reasonable target control command based on multiple candidate device action schemes.
[0073] Figure 3 A schematic flowchart illustrating the selection of a target control command from an initial control command according to an embodiment of the present invention is shown. Figure 3 As shown, step S220 may include steps S310 to S320.
[0074] In step S310, the preset mutual exclusion rules and the initial control instructions are input into the second language model to output the recognition result, wherein the recognition result indicates whether the initial control instructions include mutual exclusion instructions that conform to the preset mutual exclusion rules.
[0075] The preset mutual exclusion rules can include pre-set mutual exclusion instructions. These instructions can include triggering conditions and corresponding device actions, or they can include actions from the same device and / or actions from different devices. Mutually exclusive instructions are often control instructions that defy common sense; for example, a mutually exclusive instruction could be "increase CO2 concentration while simultaneously activating high ventilation." Activating high ventilation is highly likely to increase CO2 concentration, therefore this instruction is mutually exclusive. The second major language model can analyze the semantics of the initial control instruction and the semantics of the mutually exclusive instructions within the preset mutual exclusion rules to determine whether the initial control instruction conforms to the mutual exclusion rules.
[0076] In step S320, a target control instruction is selected from the initial control instructions based on the identification result, wherein the identification result of the target control instruction indicates that the target control instruction does not include mutually exclusive instructions.
[0077] Based on the identification results, it can be determined which initial control commands conform to mutual exclusion instructions and which do not. Therefore, target control commands can be selected only from the initial commands that do not include mutual exclusion instructions, thus avoiding mutually exclusive operations in subsequent control operations.
[0078] In the above technical solution, preset mutual exclusion rules and initial control instructions are input into a second language model to output recognition results. The recognition results indicate whether the initial control instructions include mutually exclusive instructions that conform to the preset mutual exclusion rules. Then, based on the recognition results, target control instructions are selected from the initial control instructions. The recognition result of the target control instructions indicates that the target control instructions do not include mutually exclusive instructions. This further filters the initial control instructions, preventing potential mutual exclusion instructions in the target control instructions from affecting the effectiveness of subsequent control operations. Determining more reasonable target control instructions improves the effectiveness of subsequent control operations.
[0079] For example, when the identification results of all initial control commands indicate that the initial control command includes a mutually exclusive command, a reference control command corresponding to the reference prediction result is retrieved from the historical record as the current target control command. The historical record includes the corresponding stored reference prediction results and reference control commands, and the reference prediction result is the environmental prediction result in the historical record that has the smallest deviation from the environmental prediction result.
[0080] After each target control command is determined, the environmental prediction result used to determine the target control command can be saved as a reference prediction result in the history record, and the target control command can also be saved as a reference reference control command corresponding to the reference prediction result in the history record.
[0081] Understandably, after step S310 above, the identification results of the initial control command may all indicate that the initial control command includes a mutually exclusive command. In this case, it becomes impossible to determine from the initial control command that it does not include a mutually exclusive command. Since the previously determined target control command does not contain a mutually exclusive command, the reference prediction result with the smallest deviation from the environmental prediction result can be retrieved from the historical records, and the reference control command corresponding to this reference prediction result can be determined as the current target control command. For example, the deviation between the environmental prediction result and the reference prediction result can be determined based on the differences of each environmental parameter between the environmental prediction result and the reference prediction result. The smaller this deviation, the closer the environmental prediction result and the reference prediction result are, and the more likely the reference control command corresponding to the reference prediction result is to be applicable to that reference prediction result.
[0082] In the above technical solution, when the identification results of all initial control commands indicate that the initial control command includes a mutually exclusive command, a reference control command corresponding to the reference prediction result is retrieved from the historical records and used as the current target control command. The historical records include corresponding stored reference prediction results and reference control commands, where the reference prediction result is the environmental prediction result in the historical records with the smallest deviation from the environmental prediction result. This allows for the determination of a target control command that is as reliable as possible based on historical records even when a reliable initial control command has not been determined, thus avoiding the inability to determine the target control command.
[0083] Figure 4 A schematic flowchart illustrating reinforcement training of a motion plan prediction model according to an embodiment of the present invention is shown. Figure 4 As shown, the above-mentioned greenhouse environment control method may further include steps S410 to S420.
[0084] In step S410, the number of times the step of retrieving the reference control command corresponding to the reference prediction result from the historical record is performed within a preset time window is counted.
[0085] The preset time window can be a preset time window determined based on the current time. For example, the preset time window can be a time window of a preset duration up to the current time. Whether a step operates within the preset time window can be determined by checking whether the target operation time of the step that retrieves the reference control instruction corresponding to the reference prediction result from the historical record is within the preset time window. The target operation time can be the start time, end time, or a time in between for the step.
[0086] In step S420, when the number of operations is greater than or equal to a preset threshold, the near-end strategy optimization algorithm is used to perform reinforcement learning training on the action plan prediction model based on environmental parameters and environmental prediction results.
[0087] The more times the step of retrieving the reference control command corresponding to the reference prediction result from the historical record is performed within a preset time window, the more severe the output error of the action plan prediction model. When the number of operations is greater than or equal to a preset threshold, it indicates that the output error of the action plan prediction model is too severe. In this case, the proximal policy optimization algorithm can be used to perform reinforcement learning training on the action plan prediction model to improve its performance. For example, the difference between environmental parameters and environmental prediction results can be used as a reward signal, and the proximal policy optimization algorithm can be used to perform reinforcement learning training on the action plan prediction model.
[0088] In the above technical solution, the number of operations performed within a preset time window for retrieving the reference control command corresponding to the reference prediction result from historical records is counted. Then, when the number of operations is greater than or equal to a preset threshold, a proximal policy optimization algorithm is used to perform reinforcement learning training on the action plan prediction model based on environmental parameters and environmental prediction results. This allows for timely optimization of the action plan prediction model with excessively severe errors, preventing the model from continuously outputting erroneous data. Furthermore, this reinforcement training of the action plan prediction model can prevent errors after executing control operations.
[0089] Figure 5 A schematic flowchart illustrating the selection of a target control command from an initial control command based on comprehensive indicators according to an embodiment of the present invention is shown. Figure 5 As shown, the above-mentioned greenhouse environment control method may further include step S510, and step S130 may include steps S520 to S540.
[0090] In step S510, the target execution result of each of the multiple candidate device action schemes based on environmental parameters is determined.
[0091] For each set of candidate equipment action plans, the impact of each equipment action on environmental parameters can be predetermined. For example, the impact of opening a window by 50% on carbon dioxide concentration. The impact of each equipment action on environmental parameters may be different. Therefore, based on the impact of each equipment action on environmental parameters and the environmental parameters at the current moment, the theoretical environmental parameters after executing the set of candidate equipment action plans can be comprehensively calculated as the target execution result.
[0092] In step S520, the initial control commands corresponding to each of the multiple candidate device action schemes are determined using a large language model.
[0093] Step S520 is similar to step S210, and will not be described in detail here.
[0094] In step S530, based on the target execution result of the candidate device action scheme corresponding to each initial control command and the target difference between the suitable environmental parameters of the target plants in the target greenhouse, the comprehensive index corresponding to the initial control command is calculated.
[0095] For each initial control command, the smaller the target difference, the more the environmental parameters after the execution of the candidate device action plan corresponding to that initial control command meet the growth requirements of the target plant. When the weighting coefficient corresponding to the target difference is positive, the smaller the difference, the smaller the comprehensive index corresponding to the initial control command can be. When the weighting coefficient corresponding to the target difference is negative, the larger the difference, the smaller the comprehensive index corresponding to the initial control command can be. For example, the weighting coefficient corresponding to the difference can be positive, and a comfort penalty index for the target plant can be determined based on the target difference. The comfort penalty index and the comprehensive index are positively correlated.
[0096] In addition to the difference in objectives, the comprehensive indicators corresponding to the initial control command can be calculated more comprehensively based on the device actions of the candidate device action schemes.
[0097] For example, the comprehensive index corresponding to the initial control command is also calculated based on the device energy consumption and / or device start-stop smoothness of the candidate device action scheme corresponding to the initial control command.
[0098] Equipment energy consumption can be calculated based on the equipment actions and durations of candidate equipment action schemes. For example, the total energy consumption can be calculated based on all equipment actions involved in a candidate equipment action scheme and their respective durations, and this total energy consumption can be used to calculate a comprehensive indicator. For instance, equipment energy consumption indicators can be determined based on equipment energy consumption and real-time electricity prices, and used to calculate a comprehensive indicator.
[0099] Equipment start-up and shutdown smoothness reflects the degree of change in equipment operating status corresponding to equipment actions within a certain period of time. A corresponding smoothness penalty index can be determined based on the equipment start-up and shutdown smoothness. For example, if a fan speed suddenly jumps from 0% to 100%, or a sunshade suddenly opens from 0% to 100%, the equipment start-up and shutdown smoothness is low, and the corresponding smoothness penalty index can be high. For instance, the smoothness penalty index can be determined based on parameters such as the time the equipment start-up and shutdown smoothness corresponding to the candidate equipment action scheme exceeds a preset smoothness threshold, and the number of switch state transitions.
[0100] Understandably, combining the comprehensive indicators determined by the energy consumption and / or start-stop smoothness of the candidate device action scheme corresponding to the initial control command can more comprehensively reflect the effect of the candidate device action scheme after execution and is more conducive to accurately evaluating the candidate device action scheme corresponding to the initial control command.
[0101] For example, the comprehensive index corresponding to the initial control command can be calculated according to the following formula 1:
[0102]
[0103] Where J represents the comprehensive index, H represents the number of time steps in the prediction window after the current moment for which the initial control command is applied, and C represents the number of time steps in the prediction window after the current moment. t E represents the aforementioned comfort penalty index. t The above energy consumption index, S t The smoothness penalty index mentioned above, `violations`, indicates whether the candidate device action scheme corresponding to the initial control command violates preset mandatory rules or preset logical constraints. It can take values of 0 or 1 to indicate whether preset mandatory rules or preset logical constraints are present. η, γ, and λ are constant positive numbers. Violating preset mandatory rules or preset logical constraints indicates that the candidate device action scheme corresponding to the initial control command has a conflict or is unsafe. For example, preset mandatory rules or preset logical constraints may include the aforementioned mutual exclusion rules.
[0104] In step S540, a target control command is selected from the initial control commands based on the comprehensive indicators.
[0105] Understandably, the comprehensive index at least reflects the suitability of the candidate device action schemes corresponding to the initial control command. Therefore, the initial control command with the smallest or largest comprehensive index can be selected as the most suitable target control command based on the actual situation. Taking the comprehensive index determined by Formula 1 above as an example, the initial control command with the smallest comprehensive index can be selected as the target control command.
[0106] In the above technical solution, the target execution results of multiple candidate equipment action schemes based on environmental parameters are determined. Then, using a large language model, the initial control commands corresponding to each of the candidate equipment action schemes are determined. Next, based on the target difference between the target execution result of the candidate equipment action scheme corresponding to each initial control command and the suitable environmental parameters of the target plants in the target greenhouse, a comprehensive index corresponding to the initial control command is calculated. Based on the comprehensive index, the target control command is selected from the initial control commands. In this way, the initial control commands can be comprehensively evaluated based on the comprehensive index to determine their suitability for the current situation, and thus the most suitable initial control command can be selected as the target control command.
[0107] Figure 6 A schematic flowchart illustrating the determination of target differences according to an embodiment of the present invention is shown. Figure 6 As shown, the above-mentioned greenhouse environment control may also include steps S610 to S630.
[0108] In step S610, for the target execution result of the candidate device action scheme corresponding to each initial control command, the weight coefficient corresponding to each target difference is determined according to the growth stage of the target plant. The target difference is the difference between the target execution result of the same type and the suitable environmental parameters of the growth stage of the target plant.
[0109] For example, the types of target execution results and suitable environmental parameters can be the same as the types of environmental prediction results determined in step S110 above.
[0110] The correspondence between the growth stage of the target plant and the weight coefficient corresponding to each target difference can be predetermined. Then, based on the growth stage of the target plant and this correspondence, the weight coefficient corresponding to each target difference can be determined. For example, for the flowering period and maturity period of the target plant, the corresponding weight coefficients for each target difference can be predetermined.
[0111] In step S620, the weighted difference of each target difference is calculated based on the weight coefficient and the target difference.
[0112] By multiplying each weight coefficient by its corresponding target difference, we can obtain the weighted difference of each target difference.
[0113] In step S630, based on the weighted difference of each target difference, the target difference between the target execution result of the candidate device action scheme corresponding to the initial control command and the target environmental parameters of the growth stage of the target plant is determined.
[0114] The weighted sum of the differences of each objective can be used as the objective difference.
[0115] For example, taking temperature, humidity, and carbon dioxide concentration as examples, we can determine the first target difference between the target temperature and the suitable temperature corresponding to the growth stage of the target plant in the target execution result; the second target difference between the target humidity and the suitable humidity corresponding to the growth stage of the target plant in the target execution result; and the third target difference between the target carbon dioxide concentration and the suitable carbon dioxide concentration corresponding to the growth stage of the target plant in the target execution result. Then, based on the growth stage of the target plant, we can determine the first weighting coefficient corresponding to the first target difference, the second weighting coefficient corresponding to the second target difference, and the third weighting coefficient corresponding to the third target difference. Next, we can calculate the first weighted difference, the second weighted difference, and the third weighted difference corresponding to the third target difference. Finally, we can add the first weighted difference, the second weighted difference, and the third weighted difference to obtain the target difference.
[0116] In the above technical solution, a weighting coefficient is determined for each target difference based on the growth stage of the target plant. The target differences are the differences between the execution result of the same type of target and the suitable environmental parameters. Then, based on the weighting coefficients and the target differences, a weighted difference for each target difference is calculated, and the target difference is determined based on this weighted difference. The target difference obtained in this way can more accurately assess the target difference between the execution result of the candidate device action plan corresponding to each initial control command and the suitable environmental parameters for the growth stage of the target plant.
[0117] For example, step S110 above may include step S710: inputting environmental parameters and equipment operating status into the environmental prediction model to output environmental prediction results.
[0118] Training data on environmental parameters and equipment operating status can be pre-acquired. The target environmental parameters and target equipment operating status from this training data are then input into the prediction model to output predictions of the environmental parameters within a preset future time period after the last moment of the target environmental parameters. The target environmental parameters are those collected while the target equipment operating status remains unchanged. The difference between this prediction and the target environmental parameters within the preset future time period is then calculated, and the parameters of the prediction model are adjusted using this loss value to train the model. After multiple training iterations, the environmental prediction model is obtained.
[0119] Figure 7 A schematic flowchart of a retrained environment prediction model according to an embodiment of the present invention is shown. Figure 7 As shown, the above-mentioned greenhouse environment control method may further include steps S710 to S730.
[0120] In step S720, the target execution result of each of the multiple candidate device action schemes based on environmental parameters is determined.
[0121] Step S720 is similar to step S510 above, and will not be described in detail here.
[0122] In step S730, the deviation between the actual environmental parameters and the theoretical execution result after the control operation is determined, wherein the theoretical execution result is the target execution result of the candidate device action scheme corresponding to the target control command.
[0123] Based on the correspondence between the target control command and the candidate device action plan, the theoretical execution result can be determined from the established target execution results. Then, the actual environmental parameters after executing the control operation are the actual environmental parameters after completing the candidate device action plan corresponding to the target control command. The end time of the control operation can be determined based on the runtime segment corresponding to the candidate device action plan and the start time of the control operation, and the environmental parameters at this end time can be collected as the actual environmental parameters after the control operation. Based on the difference between each type of actual environmental parameter and the corresponding theoretical execution result, the deviation between the actual environmental parameter and the theoretical execution result is determined. This deviation represents the discrepancy between the actual execution result and the expected result of the control operation. For example, the deviation can be determined based on the absolute error, mean square error, relative error, and other related errors between the actual environmental parameters and the corresponding theoretical execution result.
[0124] In step S740, when the deviation is greater than the preset deviation threshold, the parameters of the environmental prediction model are adjusted based on the preset training strategy.
[0125] When the deviation exceeds a preset deviation threshold, it indicates that the actual execution result of the control operation deviates too much from the expected result. This deviation may be due to the inaccuracy of the environmental prediction results predicted by the environmental prediction model. Therefore, the parameters of the environmental prediction model can be adjusted to improve its performance. For example, preset training strategies may include incremental updates, sliding window retraining, and parameter fine-tuning.
[0126] For example, the preset threshold can be updated based on the current time period and the distribution of the deviation over the time period.
[0127] Understandably, for the same environmental parameters, if the data collection periods are different, the changes in environmental parameters after the current moment may also differ. Therefore, for example, the environmental prediction results generated by the environmental prediction model may be biased due to the current time period, which will affect the subsequently determined action plans of candidate devices and ultimately the actual environmental parameters after the control operation is performed. For example, the prediction error for nighttime illumination is generally larger than that for daytime. Therefore, the distribution of the deviation over time periods can be predetermined, and the preset threshold in step S740 can be adjusted based on this distribution and the current time period. For example, the prediction error for illumination by the environmental prediction model is generally larger at night, so the threshold for this time period can be appropriately increased to reduce invalid corrections. In this way, the adjustment strategy of the environmental prediction model can be controlled in real time according to the current time period to reduce invalid corrections to the environmental prediction model.
[0128] In the above technical solution, the target execution results of multiple candidate device action schemes based on environmental parameters are determined, and the deviation between the actual environmental parameters after the execution of the control operation and the theoretical execution result is determined. The theoretical execution result is the target execution result of the candidate device action scheme corresponding to the target control command. Then, when the deviation exceeds a preset deviation threshold, the parameters of the environmental prediction model are adjusted based on a preset training strategy. This allows for timely adjustment of the environmental prediction model's parameters, reducing prediction errors and ensuring that the execution results of subsequent control operations better meet expectations.
[0129] Figure 8 A schematic block diagram of a greenhouse environment control device according to an embodiment of the present invention is shown. Figure 8 As shown, the greenhouse environment control device includes a first prediction module 810, a second prediction module 820, a third prediction module 830, and a control module 840.
[0130] The first prediction module 810 is used to determine the environmental prediction result of the target greenhouse in the future preset period based on the environmental parameters and equipment operating status of the target greenhouse in the preset period before the current time, wherein the equipment operating status of the target greenhouse affects the environmental parameters of the target greenhouse.
[0131] The second prediction module 820 is used to input environmental prediction results and preset knowledge rules into the action plan prediction model to obtain multiple candidate equipment action plans for the target greenhouse.
[0132] The third prediction module 830 is used to determine the target control command corresponding to one of the multiple candidate device action schemes by utilizing a large language model based on multiple candidate device action schemes.
[0133] The control module 840 is used to control the equipment in the target greenhouse according to the target control command.
[0134] According to another aspect of the present invention, an electronic device is also provided. Figure 9 A schematic block diagram of an electronic device according to an embodiment of the present invention is shown. Figure 9 As shown, the electronic device includes a processor and a memory, wherein the memory stores computer program instructions, which are executed by the processor to perform the greenhouse environment control method described above.
[0135] Furthermore, according to another aspect of the present invention, a storage medium is provided, on which program instructions are stored. When the program instructions are executed by a computer or processor, the computer or processor performs corresponding steps of the greenhouse environment control method described in the embodiments of the present invention, and is used to implement corresponding modules in the greenhouse environment control device described in the embodiments of the present invention. The storage medium may, for example, include a storage component of a tablet computer, a hard disk of a personal computer, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disc read-only memory (CD-ROM), a USB memory, or any combination of the above storage media. The computer-readable storage medium may be any combination of one or more computer-readable storage media.
[0136] According to another aspect of the present invention, a computer program product is also provided, including computer program instructions, which, when executed, are used to perform the above-described greenhouse environment control method.
[0137] Those skilled in the art can understand the specific implementation and beneficial effects of the above-described greenhouse environment control methods by reading the detailed description above. For the sake of brevity, they will not be described in detail here.
[0138] Although exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above exemplary embodiments are merely illustrative and are not intended to limit the scope of the invention thereto. Various changes and modifications can be made therein by those skilled in the art without departing from the scope and spirit of the invention. All such changes and modifications are intended to be included within the scope of the invention as claimed in the appended claims.
[0139] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0140] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.
[0141] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.
[0142] Similarly, it should be understood that, in order to streamline the invention and aid in understanding one or more of the various aspects of the invention, features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the description of exemplary embodiments of the invention. However, this approach should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the corresponding claims, its inventive point lies in solving the corresponding technical problem with fewer features than all of those in a single disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the invention.
[0143] Those skilled in the art will understand that, apart from the mutual exclusion of features, all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or apparatus so disclosed can be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.
[0144] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features but not others included in other embodiments, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the claims, any of the claimed embodiments can be used in any combination.
[0145] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some modules in a greenhouse environment control device according to embodiments of the present invention. The present invention can also be implemented as an apparatus program (e.g., a computer program and computer program product) for performing some or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.
[0146] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.
[0147] The above description is merely a specific embodiment of the present invention or an explanation of that embodiment. The scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A method for controlling greenhouse environment, characterized in that, The method includes: Based on the environmental parameters and equipment operating status of the target greenhouse during a preset period prior to the current moment, the environmental prediction result of the target greenhouse during a future preset period is determined, wherein the equipment operating status of the target greenhouse affects the environmental parameters of the target greenhouse; The environmental prediction results and preset knowledge rules are input into the action plan prediction model to obtain multiple candidate equipment action plans for the target greenhouse. Using a large language model, based on the multiple sets of candidate device action schemes, a target control command corresponding to one of the multiple sets of candidate device action schemes is determined; Control operations are performed on the equipment of the target greenhouse according to the target control command.
2. The method according to claim 1, characterized in that, The large language model includes a first large language model and a second large language model. The step of using a large language model to determine a target control command corresponding to one of the multiple candidate device action schemes includes: The multiple candidate device action schemes are respectively input into the first large language model to output their respective initial control commands; Using the second major language model, the target control instruction is determined from the initial control instruction.
3. The method according to claim 2, characterized in that, The step of using the second large language model to determine the target control command from the initial control command includes: The preset mutual exclusion rules and the initial control instructions are input into the second large language model to output the recognition result, wherein the recognition result indicates whether the initial control instructions include mutual exclusion instructions that conform to the preset mutual exclusion rules; Based on the identification result, a target control instruction is selected from the initial control instructions, wherein the identification result of the target control instruction indicates that the target control instruction does not include mutually exclusive instructions.
4. The method according to claim 3, characterized in that, The method further includes: When the identification results of all initial control commands indicate that the initial control command includes a mutual exclusion command, a reference control command corresponding to the reference prediction result is retrieved from the historical record as the current target control command. The historical record includes corresponding stored reference prediction results and reference control commands, and the reference prediction result is the environmental prediction result with the smallest deviation from the environmental prediction result in the historical record.
5. The method according to claim 4, characterized in that, The method further includes: The number of operations performed within a preset time window is counted for retrieving reference control instructions corresponding to reference prediction results from historical records. When the number of operations is greater than or equal to a preset threshold, the action plan prediction model is trained by reinforcement learning using a proximal policy optimization algorithm based on the environmental parameters and the environmental prediction results.
6. The method according to claim 2, characterized in that, The step of inputting the multiple candidate device action schemes into the first large language model to output their respective initial control commands includes: The multiple candidate device action schemes are input into the first large language model to output their respective initial control commands and natural language descriptions.
7. The method according to claim 1, characterized in that, The method further includes: Determine the target execution result of each of the multiple candidate device action schemes based on the environmental parameters; The step of using a large language model to determine a target control command corresponding to one of the multiple candidate device action schemes includes: Using the large language model, the initial control commands corresponding to each of the multiple candidate device action schemes are determined; Based on the target execution result of the candidate device action scheme corresponding to each initial control command and the target difference between the suitable environmental parameters of the target plants in the target greenhouse, the comprehensive index corresponding to the initial control command is calculated. Based on the comprehensive indicators, the target control command is selected from the initial control commands.
8. The method according to claim 7, characterized in that, The calculation of the comprehensive index corresponding to the initial control command is also based on the equipment energy consumption and / or equipment start-stop smoothness of the candidate equipment action scheme corresponding to the initial control command.
9. The method according to claim 7, characterized in that, The method further includes: The target execution result for each candidate device action scheme corresponding to the initial control command. Based on the growth stage of the target plant, determine the weight coefficient corresponding to each target difference, where the target difference is the difference between the execution result of the same type of target and the suitable environmental parameters of the growth stage of the target plant. Calculate the weighted difference of each target difference based on the weighting coefficients and the target differences; Based on the weighted difference of each target difference, the target difference between the target execution result of the candidate device action scheme corresponding to the initial control command and the target environmental parameters of the suitable growth stage of the target plant is determined.
10. The method according to claim 1, characterized in that, The step of determining the environmental prediction result of the target greenhouse for a future preset period based on the environmental parameters and equipment operating status of the target greenhouse during a preset period prior to the current moment includes: The environmental parameters and the equipment operating status are input into the environmental prediction model to output the environmental prediction results; The method further includes: Determine the target execution result of each of the multiple candidate device action schemes based on the environmental parameters; Determine the deviation between the actual environmental parameters after executing the control operation and the theoretical execution result, wherein the theoretical execution result is the target execution result of the candidate device action scheme corresponding to the target control command; When the deviation exceeds a preset deviation threshold, the parameters of the environment prediction model are adjusted based on a preset training strategy.
11. The method according to claim 10, characterized in that, The method further includes: The preset threshold is updated based on the current time period and the distribution of the deviation over the time period.
12. A greenhouse environment control device, characterized in that, include: The first prediction module is used to determine the environmental prediction result of the target greenhouse in the future preset period based on the environmental parameters and equipment operating status of the target greenhouse in the preset period before the current time, wherein the equipment operating status of the target greenhouse affects the environmental parameters of the target greenhouse; The second prediction module is used to input the environmental prediction results and preset knowledge rules into the action plan prediction model to obtain multiple candidate equipment action plans for the target greenhouse. The third prediction module is used to determine a target control command corresponding to one of the multiple candidate device action schemes based on the large language model. The control module is used to perform control operations on the equipment of the target greenhouse according to the target control command.
13. An electronic device comprising a processor and a memory, characterized in that, The memory stores computer program instructions, which, when executed by the processor, are used to perform the greenhouse environment control method as described in any one of claims 1 to 11.
14. A storage medium on which program instructions are stored, characterized in that, The program instructions, when executed, are used to perform the greenhouse environment control method as described in any one of claims 1 to 11.
15. A computer program product comprising computer program instructions, characterized in that, The computer program instructions, when executed, are used to perform the greenhouse environment control method as described in any one of claims 1 to 11.