Environment self-control seedling greening greenhouse and multi-dimensional parameter scheduling method thereof

Abstract

The application discloses an environment self-control seedling greening greenhouse and a multi-dimensional parameter scheduling method thereof, and relates to the automatic control technology of greenhouse environment. The environment controller of the greenhouse fuses indoor and outdoor environment data according to the credibility of a sensing node, identifies sunrise and sunset events according to outdoor illumination, generates a light period phase and a daily cumulative effective light quantity, generates a dynamic environment target domain containing an allowable temperature interval and an air water vapor partial pressure target in combination with seedling growth stages and carbon dioxide concentration, and obtains an environment deviation vector according to the dynamic environment target domain. Then, the environment controller predicts environment changes of each candidate action combination according to an on-line self-correcting actuator environment action fingerprint, eliminates combinations with mutually cancelled actions, and selects and outputs a combination with the minimum scheduling cost to a skylight, a side window, a warming mechanism and an automatic irrigation mechanism. Therefore, the environment target continuously changes with the actual light environment, and mutual cancellation and frequent start and stop of equipment actions are reduced.

Description

Technical Field

[0001] This invention relates to automatic control technology for greenhouse environment, and in particular to an environmentally self-controlled seedling greening greenhouse and its multi-dimensional parameter scheduling method. Background Technology

[0002] Rice factory-style seedling production lines typically follow a process flow of cold seed storage, seed pretreatment, hot water disinfection, seed soaking and germination, automatic sowing, and germination in a germination chamber. After germination, the seedling trays are then transferred to a green greenhouse for centralized cultivation. Seedling cultivation is a crucial, continuous process linking the preceding and following stages of seedling raising. It involves a large number of seedling trays, high density, concentrated production rhythm, and rapid changes in the indoor environment due to day-night cycles and sudden weather changes, significantly differing from the long-term cultivation in ordinary vegetable greenhouses. The control objective of a green greenhouse is not to maintain a constant temperature or humidity, but rather to adjust temperature, humidity, ventilation, and irrigation strategies in real time according to the seedling growth stage, indoor and outdoor environment, natural light, and equipment operating status.

[0003] Most current green greenhouses are equipped with two sets of sensors: an indoor sensor and an outdoor sensor. The indoor sensor detects indoor temperature, relative humidity or air moisture content, light intensity, and carbon dioxide concentration. The outdoor sensor detects outdoor temperature, outdoor humidity, outdoor light intensity, rainfall, and wind speed. A field controller can connect to several sensor nodes and drive automatic irrigation, skylights, side windows, heating equipment, and internal circulation or ventilation equipment. The corresponding human-machine interface typically has functions such as alarm setting, time interval management, temperature management, humidity management, and internal linkage settings. For example, it can identify sunrise and sunset based on light thresholds and set time offsets before sunrise, after sunrise, before sunset, and after sunset; it can set minimum target temperature, temperature change rate, and brightness change coefficient (ILR); it can set different maximum target temperatures according to different carbon dioxide concentration ranges; humidity control can select between saturated vapor pressure differential control and relative humidity control; and each actuator can operate according to time intervals.

[0004] While the aforementioned functions are relatively comprehensive in terms of parameter types, their control structure still relies on a set of upper and lower limits for each environmental parameter, with each device operating according to fixed rules. Natural light, carbon dioxide concentration, temperature, and air moisture levels are not independent of each other. The same 25°C has different implications for seedlings at the beginning of natural light, during strong light, and after sunset. Similarly, the same 80% relative humidity corresponds to different actual air moisture content and saturated vapor pressure differential at different temperatures. Fixed time intervals cannot adapt to seasonal changes or daily weather variations, requiring operators to frequently adjust the start and end times of skylights, side windows, heating, and irrigation. Control targets generated based on independent setpoints exhibit abrupt changes during sudden weather events, easily maintaining higher target temperatures under weak light and low carbon dioxide concentration conditions as if under strong light, leading to unnecessary adjustments. Switching between relative humidity control and saturated vapor pressure differential control can cause sudden changes in device behavior due to the altered definition of the control quantity.

[0005] On the other hand, the various actuators in a green greenhouse simultaneously affect multiple environmental parameters. Opening skylights or side windows lowers indoor temperature and humidity while altering indoor carbon dioxide concentration; the direction of this effect also depends on the indoor-outdoor environmental gradient. Heating increases temperature while changing relative humidity and saturated vapor pressure difference. Automatic irrigation meets the water requirements of the seedling trays while increasing local humidity. If temperature, humidity, and carbon dioxide each form an independent closed loop, under low-temperature, high-humidity conditions, the temperature closed loop may require heating while the humidity closed loop simultaneously requires large-scale ventilation. These two actions cancel each other out in the temperature dimension, resulting in inefficient energy consumption. Furthermore, multiple closed loops repeatedly compete for control of the same actuator, leading to frequent start-ups and shutdowns. In addition, field sensor nodes experience drift, disconnections, and communication quality degradation during long-term operation. If erroneous measurements directly enter the closed loop, these contradictions will be further amplified.

[0006] Therefore, the core technical problem faced by the existing green greenhouse control method is that each execution device is driven separately with fixed time periods and independent set values, so that the environmental target cannot be continuously generated according to the actual light environment and multi-parameter coupling state of the day; and the multi-dimensional coupling effect of each execution device is not uniformly evaluated, so the device actions cancel each other out and start and stop frequently. Summary of the Invention

[0007] The purpose of this invention is to provide an environmentally controlled seedling greening greenhouse and its multi-dimensional parameter scheduling method, so that the environmental targets of the greening greenhouse can be continuously generated according to the actual measured light environment, seedling growth stage and carbon dioxide conditions of the day, and each execution device can be uniformly selected according to its actual multi-dimensional role in the current environment, thereby reducing mutual cancellation between device actions and frequent start-stop.

[0008] This invention provides an environmentally controlled seedling greenhouse, comprising a greenhouse body, a seedling greening area within the greenhouse body, an indoor environmental sensing unit, an outdoor environmental sensing unit, a communication unit, an environmental actuator, and an environmental controller connected to the indoor and outdoor environmental sensing units via the communication unit and connected to the environmental actuator. The environmental actuator includes at least two of the following: a skylight, a side window, a heating mechanism, and an automatic irrigation mechanism. The environmental controller is configured to perform the following operations according to a scheduling cycle:

[0009] For similar sensing nodes, the consistency of their time change rate, spatial deviation, and consistency of physical quantities are calculated respectively. Based on this, the reliability of each sensing node is determined. The similar detection data are then weighted and fused according to the reliability to obtain environmental state quantities including indoor temperature, indoor air moisture state, indoor carbon dioxide concentration, indoor light intensity, outdoor temperature, outdoor air moisture state, and outdoor light intensity.

[0010] The sunrise and sunset times for the day are determined based on the outdoor light intensity, a preset light threshold, and the duration of the light condition. The results for the day are constrained by the average of the most recent N days. The daytime is normalized using the sunrise and sunset times as endpoints to obtain the photoperiod phase. The cumulative effective light intensity for the day is obtained by integrating the outdoor light intensity from the sunrise time to the current time. The preset time adjustments before sunrise, after sunrise, before sunset, and after sunset are used as offsets relative to the sunrise and sunset times.

[0011] The dynamic environmental target domain is generated by using the seedling growth stage, the photoperiod phase, the cumulative effective light intensity of the day, and the indoor carbon dioxide concentration as scheduling coordinates. The dynamic environmental target domain includes the allowable temperature range defined by the lower limit of the target temperature and the upper limit of the target temperature, as well as the target water vapor state expressed by the actual water vapor partial pressure of the air. The indoor temperature and the indoor air moisture state are uniformly converted into the actual water vapor partial pressure of the air as a common intermediate variable for humidity control.

[0012] Based on the environmental state quantity and the dynamic environmental target domain, calculate an environmental deviation vector containing temperature, air moisture, and carbon dioxide components.

[0013] The actuator environmental action fingerprint of each actuator is stored. The actuator environmental action fingerprint characterizes the direction and intensity of the actuator’s effect on temperature, air moisture state and carbon dioxide concentration per unit amount of action. The action direction of the skylight and side window is determined online based on the difference between indoor and outdoor temperature and the difference between indoor and outdoor air moisture state.

[0014] Candidate action combinations are constructed based on the discrete action levels of each actuator. The environmental state after execution of each candidate action combination is predicted based on the actuator environmental action fingerprint, and the scheduling cost is calculated. After eliminating candidate action combinations whose main action directions cancel each other out, the one with the lowest scheduling cost is selected and output to the corresponding actuator. After the actuator completes the action and the response delay time of the actuator has elapsed, the actuator environmental action fingerprint is updated with weight by the ratio of the change in the environmental state quantity to the product of the corresponding action quantity and the response delay time.

[0015] As a preferred embodiment of the present invention, the indoor environment sensing unit includes multiple sensing nodes, which are respectively arranged at different locations in the seedling greening area and near the skylight and the air outlet of the heating mechanism, with the installation height located near the seedling canopy. The environment controller is also configured to reduce the weight of a sensing node in the weighted fusion when the confidence level of the sensing node is lower than a first threshold, to remove its detection data and output an alarm when the confidence level of the sensing node is lower than a second threshold, to maintain the environmental state quantity of the previous scheduling cycle and set a degradation flag when the sum of the confidence levels of similar sensing nodes is lower than a set lower limit, and to stop humidity scheduling based on saturated water vapor pressure difference and only retain temperature safety control when the indoor air moisture state fails, to prohibit the use of the highest allowable temperature corresponding to the higher carbon dioxide concentration range when the indoor carbon dioxide concentration fails, and to generate the photoperiodic phase by replacing the measured value of the day with the average of the sunrise and sunset event times of the most recent N days when the outdoor light intensity fails.

[0016] As a preferred embodiment of the present invention, the environmental actuator further includes a circulating fan installed above the seedling greening area and an exhaust fan installed on the main gable wall of the greenhouse. The skylight and side window are respectively equipped with opening feedback elements. The discrete action positions of each actuator include a closed position and at least one action position. The environmental controller is also configured to limit the opening position of the skylight according to the outdoor rainfall and outdoor wind speed and to eliminate candidate action combinations that violate equipment safety conditions. When calculating the scheduling cost, for the actuators that have been activated in the previous scheduling cycle but have not reached the minimum holding time, the start-stop cost item corresponding to the change of their position is increased.

[0017] This invention also provides a multi-dimensional parameter scheduling method for an environmentally controlled seedling greening greenhouse as described above, comprising the following steps:

[0018] S1. The environmental controller collects the detection data of the indoor and outdoor environmental sensing units according to the scheduling cycle. For similar sensing nodes, the consistency of their time change rate, spatial deviation, and physical quantity consistency are calculated respectively. Based on this, the reliability of each sensing node is determined. The detection data of similar types are weighted and fused according to the reliability to obtain environmental state quantities including indoor temperature, indoor air moisture state, indoor carbon dioxide concentration, indoor light intensity, outdoor temperature, outdoor air moisture state, and outdoor light intensity. This suppresses the detection values ​​of drifting or failed nodes before they participate in the control.

[0019] S2. Determine the sunrise and sunset times for the day based on the outdoor light intensity, a preset light threshold, and the duration of the light condition. Apply a deviation constraint to the daily determination result using the average of the sunrise and sunset times over the most recent N days. Normalize the daytime using the sunrise and sunset times as endpoints to obtain the photoperiod phase representing the current position within the day's photoperiod. Integrate the outdoor light intensity from the sunrise time to the current time to obtain the cumulative effective light intensity for the day. Interpret the preset time adjustments before sunrise, after sunrise, before sunset, and after sunset as offsets relative to the sunrise and sunset times, forming a device control range that drifts with the day's light environment.

[0020] S3. A dynamic environmental target domain is generated using the seedling growth stage, the photoperiod phase, the cumulative effective light intensity of the day, and the indoor carbon dioxide concentration as scheduling coordinates. The dynamic environmental target domain includes an allowable temperature range defined by the lower limit of the target temperature and the upper limit of the target temperature, as well as the target water vapor state expressed by the actual water vapor partial pressure of the air. The lower limit of the target temperature is determined by the reference temperature corresponding to the seedling growth stage and the correction amount of the photoperiod phase. The upper limit of the target temperature is taken as the smaller value between the highest allowable temperature corresponding to the carbon dioxide concentration range and the sum of the lower limit of the target temperature, the normalized value of the light intensity, and the correction amount of the cumulative effective light intensity. The change in the target temperature between adjacent scheduling cycles is limited by a preset temperature change rate, so that the target temperature moves continuously rather than changes abruptly.

[0021] S4. Convert the indoor temperature and indoor air moisture state into the actual water vapor partial pressure of the air, and use the actual water vapor partial pressure of the air as the common intermediate variable for humidity control; when switching between relative humidity as the humidity control coordinate and saturated water vapor pressure difference as the humidity control coordinate, keep the target water vapor state unchanged, and use the switched humidity control coordinate to back-calculate the switched humidity control target value based on the saturated water vapor pressure corresponding to the current indoor temperature, so that the control target before and after the switch corresponds to the same air moisture state;

[0022] S5. Calculate an environmental deviation vector containing temperature, air moisture, and carbon dioxide components based on the environmental state quantity and the dynamic environmental target domain.

[0023] S6. Establish an actuator environmental action fingerprint for each actuator. The actuator environmental action fingerprint characterizes the direction and intensity of the actuator's action on temperature, air moisture state and carbon dioxide concentration under a unit amount of action. The action direction of the skylight and side window is determined online based on the difference between indoor and outdoor temperature and the difference between indoor and outdoor air moisture state.

[0024] S7. Construct candidate action combinations according to the discrete action levels of each actuator, predict the environmental state after each candidate action combination is executed based on the actuator environmental action fingerprint, and calculate the scheduling cost including the predicted environmental deviation item, resource usage cost item, start-stop cost item and conflict cost item. After eliminating candidate action combinations whose main action directions cancel each other out, select the candidate action combination with the smallest scheduling cost as the execution instruction output for this scheduling cycle.

[0025] S8. After the actuator completes the action and the response delay time of the actuator has elapsed, the change in the environmental state quantity is collected. The actuator environmental action fingerprint is updated by weighting the ratio of the change quantity to the product of the corresponding action quantity and the response delay time, and then the process returns to step S1.

[0026] As a preferred technical solution of the present invention, in step S2, when the outdoor light intensity exceeds the preset sunrise light threshold and the duration reaches the first set duration, the moment of exceeding the threshold is determined as the sunrise event moment of the day; when the outdoor light intensity falls below the preset sunset light threshold and the duration reaches the second set duration, the moment of falling below the threshold is determined as the sunset event moment of the day; when the deviation between the determined moment and the corresponding average of the most recent N days exceeds the allowable range, the sunrise event moment or sunset event moment of the day is not updated, but the light cycle phase is determined by the average value and the current continuous light duration, so as to avoid false triggering caused by short-term light drop in cloudy and rainy weather.

[0027] As a preferred embodiment of the present invention, in step S3, the correction amount of the photoperiod phase increases as the photoperiod phase progresses from sunrise to noon and decreases as it progresses from noon to sunset; the correction amount of the cumulative effective light intensity is taken as a non-positive value when the cumulative effective light intensity of the day is lower than the set light intensity corresponding to the seedling growth stage; the maximum allowable temperature corresponding to the indoor carbon dioxide concentration in the lower concentration range is lower than the maximum allowable temperature corresponding to the higher concentration range; when the obtained target temperature upper limit is lower than the target lower limit, the target temperature upper limit is taken to be equal to the target lower limit.

[0028] As a preferred technical solution of the present invention, in step S4, when the indoor temperature is lower than the set temperature, is in the nighttime interval after the sunset event, or the indoor temperature change rate is greater than the set value, relative humidity is selected as the humidity control coordinate; when it is in the daytime interval after the sunrise event and the indoor light intensity is greater than the set light value, saturated water vapor pressure difference is selected as the humidity control coordinate; and the condensation risk margin is determined based on the difference between the indoor temperature and the dew point temperature determined by the actual water vapor partial pressure of the air. When the condensation risk margin is less than the set threshold, the weighting coefficient of the air water vapor component in the scheduling cost of step S7 is increased.

[0029] As a preferred technical solution of the present invention, in step S6, the sign of the temperature component of the actuator environmental action fingerprint of the skylight and side window is consistent with the sign of the difference between the outdoor temperature and the indoor temperature, and the sign of its air moisture component is consistent with the sign of the difference between the outdoor air moisture state and the indoor air moisture state. Moreover, its effect intensity increases with the increase of the absolute value of the corresponding difference and the opening degree. The temperature component in the actuator environmental action fingerprint of the heating mechanism is positive, and its air moisture component is determined according to the change in relative humidity and saturated vapor pressure difference caused by the change in saturated vapor pressure after the indoor temperature rises. The air moisture component in the actuator environmental action fingerprint of the automatic water filling mechanism is positive. In step S8, the weighted update is performed only when the change amplitude of the action amount is greater than the set identifiable threshold and the credibility of the sensor node involved in the calculation is not lower than the first threshold. The weighted update performs a first-order weighting on the current identification result and the historical actuator environmental action fingerprint according to the forgetting coefficient. The forgetting coefficient is greater than 0 and less than 1, and the updated actuator environmental action fingerprint is limited to a set multiple range of its initial value.

[0030] As a preferred technical solution of the present invention, in step S7, an actuator conflict diagram is established with each actuator as a node. When the actuator environmental action fingerprints of two actuators have opposite action directions in the principal component direction of the environmental deviation vector, and the action intensity of both in this direction is greater than the set ratio of the required change in this direction, a conflict edge is established between the two, and candidate action combinations that simultaneously include the high-level actions of the actuators at both ends of the conflict edge are eliminated. When the environmental deviation vector simultaneously represents the need to increase the temperature and decrease the air moisture state, the conflict edge between the heating mechanism and the sunroof or side window is released, allowing the action of the heating mechanism to be combined with the low opening level of the sunroof or side window.

[0031] As a preferred technical solution of the present invention, it further includes converting external management instructions and internal linkage tasks into task requests containing task type, priority, duration and latest completion time, and adding the task requests as constraints to the candidate action combination construction in step S7. When there is a device safety alarm, the device is in a testing state or the condensation risk margin is less than a set threshold, the task request corresponding to the automatic watering mechanism is postponed until the constraint is lifted or the latest completion time is reached.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] A photoperiodic phase is established using sunrise and sunset events measured outdoors as endpoints. The average of the most recent N days is used to constrain the daily determination result, transforming the time adjustments in the human-machine interface—originally given as clock time—into offsets relative to the actual photoperiodic events of the day. As the seasons change and daily actual sunshine duration varies, the action ranges for skylights, side windows, heating, and water filling shift accordingly. Operators do not need to reset these settings daily, and short-term drops in sunshine due to cloudy or rainy weather are not immediately recognized as sunset events.

[0034] The seedling growth stage, photoperiod phase, cumulative effective light intensity, and carbon dioxide concentration are used together to generate a dynamic environmental target domain, replacing the separately given temperature and humidity setpoints. A rate-of-change constraint is imposed on the target temperature change in adjacent scheduling cycles, allowing the target temperature to move continuously. Therefore, under weak light or low carbon dioxide concentration conditions, the target temperature is no longer maintained at a higher level as under strong light conditions, reducing adjustments caused by unreasonable target values.

[0035] Using the actual water vapor partial pressure of the air as a common intermediate variable for humidity control, an equivalent conversion is performed between relative humidity control and saturated water vapor pressure difference control. When switching control coordinates, the moisture target is not regenerated, but the target value used for display and adjustment is calculated from the current water vapor state target. This ensures that the control target before and after the switch corresponds to the same air moisture state, reducing the sudden changes in equipment action when the control quantity definition changes.

[0036] An actuator environmental action fingerprint is established for each actuator, representing the direction and intensity of its action on temperature, air moisture state, and carbon dioxide concentration. This fingerprint is then updated with weighted values ​​based on environmental changes collected after the response delay following the action. This allows skylights and side windows to be identified as having different effects under different seasons and indoor / outdoor gradients. During scheduling, an environmental deviation vector is first generated, and then the equipment combination is selected in reverse. A conflict graph is used to eliminate combinations where the main action directions cancel each other out. Combined with start-stop cost terms and minimum hold time, this reduces ineffective operating conditions such as large-scale ventilation and continuous heating coexisting, and also reduces frequent start-stops caused by multiple independent closed loops competing for the same actuator.

[0037] The reliability of sensor nodes is determined based on the consistency of time rate of change, spatial deviation, and consistency among physical quantities, and then weighted and fused accordingly. This suppresses low-reliability data before it participates in scheduling. When critical sensor data fails, it is transferred to more stringent degradation control instead of directly using erroneous measurements, thus improving the continuous operation capability under abnormal field conditions such as sensor drift and node disconnection. Attached Figure Description

[0038] Figure 1 This is a flowchart illustrating a preferred embodiment of the multi-dimensional parameter scheduling method of the present invention;

[0039] Figure 2 This is a schematic diagram of the structure of an environmentally controlled seedling greening greenhouse according to a preferred embodiment of the present invention.

[0040] Attached reference numerals: 1. Greenhouse main body; 2. Seedling greening area; 3. Indoor environment sensing unit; 4. Outdoor environment sensing unit; 5. Environment controller; 6. Communication unit; 7. Skylight; 8. Side window; 9. Heating mechanism; 10. Automatic watering mechanism; 11. Circulating fan; 12. Exhaust equipment. Detailed Implementation

[0041] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of protection of the present invention.

[0042] like Figure 2 As shown, the environmentally controlled seedling greening greenhouse of this embodiment includes a greenhouse body 1, a seedling greening area 2 set within the greenhouse body 1, an indoor environmental sensing unit 3, an outdoor environmental sensing unit 4, an environmental controller 5, a communication unit 6, and an environmental actuator. The environmental actuator includes skylights 7 installed on both sides of the roof ridge of the greenhouse body 1, side windows 8 installed on the side walls of the greenhouse body 1, a heating mechanism 9, an automatic irrigation mechanism 10 laid above the seedling greening area 2, a circulating fan 11 suspended above the seedling greening area 2, and an exhaust device 12 installed on the gable wall of the greenhouse body 1. The greenhouse body 1 is a multi-span film greenhouse with a span of 8m, a bay width of 4m, an eaves height of 3.2m, a total of 4 spans and 6 bays, covering an area of ​​approximately 768m². Movable seedbeds are arranged in seedling greening area 2. The seedbed surface is 0.7m high. Seedling trays transferred from the germination room are received according to the rhythm of the seedling production line. The seedling trays are 58cm×28cm in size. Five to six trays are placed per unit area. The stay time of a single batch of seedling trays in seedling greening area 2 is 3 to 7 days.

[0043] The indoor environment sensing unit 3 includes 8 sensing nodes, of which 4 are combined temperature and humidity nodes, 2 are light nodes, and 2 are carbon dioxide nodes. These are located at the four corners, center, and near the skylight 7 and the air outlet of the heating mechanism 9 in the seedling greening area 2, respectively. The installation height is 30cm above the seedbed surface, near the seedling canopy. The outdoor environment sensing unit 4 is installed on a weather support 2m above the roof of the greenhouse main body 1, and includes an outdoor temperature and humidity sensor, a total radiation sensor, a rain gauge switch, and a wind speed sensor. Both the indoor environment sensing unit 3 and the outdoor environment sensing unit 4 are connected to the environment controller 5 via an RS485 bus through the communication unit 6. The communication protocol uses Modbus RTU with a baud rate of 9600bps. Each sensing node occupies slave addresses 1 to 8, while the outdoor environment sensing unit 4 occupies slave address 9. The environmental controller 5 is an embedded controller, equipped with 32 relay outputs, 8 analog outputs ranging from 4mA to 20mA, and a power-off retention memory of no less than 4MB. Its outputs are connected to the gear and rack drive motor of the skylight 7, the film rolling motor of the side window 8, the hot air furnace control circuit of the heating mechanism 9, the solenoid valve of the automatic water filling mechanism 10, the contactor of the circulating fan 11, and the exhaust equipment 12. The travel time from fully closed to fully open for the skylight 7 is 180s, and for the side window 8 it is 120s. Both are equipped with opening feedback potentiometers. The heating mechanism 9 has two power settings: low power and high power. The automatic water filling mechanism 10 is divided into two water filling zones, with a single zone irrigation flow rate of 1.2m³ / h.

[0044] like Figure 1 As shown, the sampling period of the environmental controller 5 is 10s, and the scheduling period Δt is 60s. In each scheduling period, the following steps are executed sequentially: environmental perception and reliable fusion (S1), optical periodic phase generation (S2), dynamic environmental target domain generation (S3), humidity control coordinate unification (S4), environmental deviation decomposition (S5), actuator action inversion (S6), conflict arbitration and command output (S7), and effect write-back (S8). The output of the previous step constitutes the constraint of the next step. All calculations are completed in a single scheduling period, and the measured time does not exceed 200ms.

[0045] In step S1, the environmental controller 5 polls each sensor node of the indoor environmental sensing unit 3 and the outdoor environmental sensing unit 4 every 10 seconds. A single communication timeout is defined as 300ms; if three consecutive reading failures occur, the node is marked as having interrupted communication. The raw data is first filtered by a 5-point median filter to remove pulse interference, and then filtered by a first-order low-pass filter with a time constant of 30s. The filtered sampled values ​​are used in subsequent reliability calculations. For the i-th sensor node, its reliability is calculated according to... Synthesis, in this embodiment , , The values ​​are set to 0.3, 0.4, and 0.3 respectively. The time consistency component is calculated as follows: Calculation, where The sampling interval is... The maximum allowable rate of change for this physical quantity in a greenhouse environment is defined as follows: indoor temperature 1℃ / min, relative humidity 10% / min, carbon dioxide concentration 200 μmol / (mol·min), and light intensity 500 W / (m²·min). The spatial consistency component is calculated as follows: Calculation, where The median of the detection values ​​for the same type of nodes, with allowable bias bandwidth. The indoor temperature was set to 2℃, the relative humidity to 8%, and the carbon dioxide concentration to 100 μmol / mol. The initial value of the physical consistency component was set to 1.0, and each physical quantity was verified according to its fundamental relationship with the others. Verification conditions included: the dew point temperature calculated from the relative humidity and indoor temperature at that node must not exceed the indoor temperature; the light intensity during the nighttime interval must not continuously exceed 10 W / m²; and the carbon dioxide concentration must not decrease rapidly when both skylight 7 and side window 8 are closed and there is no water injection. For each condition not met, 0.3 was immediately subtracted from the initial value, and the final result was limited to between 0 and 1. To avoid fluctuations in the reliability itself, the environmental controller 5 performed a first-order filter on the synthesized result. Fusion value according to The calculation process is as follows: when the confidence level of a sensor node falls below the first threshold of 0.6, its weight automatically decreases according to its confidence level. When it falls below the second threshold of 0.3 for five consecutive scheduling cycles, the environmental controller 5 removes the sensor node from the fusion process and outputs an alarm via the communication unit 6. The confidence level of a node whose communication is interrupted is directly set to 0. To ensure that the above fusion operation still converges under extreme conditions, the environmental controller 5 first performs a validity check on the denominator before calculation. When the sum of the confidence levels of nodes of the same type is lower than the set lower limit of 0.05, the division operation is not performed. Instead, the sampling result of this time is discarded, the environmental state quantity of the previous scheduling cycle is kept, and the degradation flag of the physical quantity is set, while the degradation timer is started. When there is only one node of the same type with a confidence level not lower than the first threshold, the detection value of that node is directly used as the fusion result, and the degradation flag is also set. After the above processing, the indoor temperature, indoor relative humidity, indoor carbon dioxide concentration, indoor light intensity, outdoor temperature, outdoor relative humidity, outdoor light intensity, rainfall status, and wind speed are obtained, which constitute the environmental state quantities of this scheduling cycle.

[0046] In step S2, the environmental controller 5 generates photoperiod coordinates based on the fused outdoor light intensity. Sunrise events are detected within a daily window from 03:00 to 12:00. When the outdoor light intensity exceeds the sunrise light threshold of 50 W / m² and remains above it for more than 10 minutes, this point of exceedance is recorded as the sunrise event time for that day. Sunset events are detected within the daily window from 15:00 to 23:00. When the outdoor light intensity falls below the sunset light threshold of 30W / m² for more than 10 minutes, the moment of this fall is recorded as the sunset event time for that day. The environmental controller 5 stores the most recent N days of valid sunrise and sunset data in the power-off retention storage area, where N is 5, and calculates... and If the deviation between the event time detected on a given day and the corresponding mean exceeds the allowable range of 60 minutes, the event time for that day will not be updated. Instead, the mean and the current duration of continuous sunlight will be used together for judgment. For example, if a thunderstorm occurs in the afternoon and causes the outdoor light intensity to drop to 20 W / m² for 15 minutes, the event time will be considered as a normal value. Since the event occurred more than 60 minutes earlier than expected, it is not considered a sunset event, and the photoperiod phase will continue to increase based on the mean. After the light recovers, the measured value will still be used in subsequent statistics. The photoperiod phase during the daytime is calculated as follows: Calculations show that the daytime conditions are met. ; At night, the same normalization is applied from the time of sunset on the current day to the predicted time of sunrise on the next day, and the values ​​satisfy the following: This ensures that day and night form a continuous coordinate system without a jump at midnight. The cumulative effective light intensity for the day is calculated as follows: Accumulated over cycles, among which The combined outdoor light intensity is expressed in W / m². The result of the summation is expressed in seconds (s), and the unit of the summation is J / m². Multiply by... Then converted to MJ / m², The value is reset to zero at sunrise each day, and the accumulated value from the previous day is carried over. (Based on a test on a certain day...) It is 05:40. Taking 19:10 as an example, the current time 10:00 corresponds to Based on this, the settings in the human-computer interface, such as 90 minutes before sunrise, 120 minutes after sunrise, 60 minutes before sunset, and 60 minutes after sunset, which were originally given as clock times, were interpreted as relative to... and The offset of the skylight 7, side window 8, heating mechanism 9, and automatic irrigation mechanism 10 automatically drifts according to the daytime sunlight environment. The seedling growth stage variable G is selected by the operator on the human-machine interface, or automatically calculated from the moment the seedling trays enter the seedling greening area 2. In this embodiment, it is divided into the initial greening period (0-2 days after entering the trays), the middle greening period (2-5 days), and the hardening-off period (after 5 days), and is correlated with... , Together they form the scheduling coordinates .

[0047] In step S3, the environmental controller 5 generates a dynamic environmental target domain based on the scheduling coordinates. The target temperature lower limit is set according to... Determined, including the reference temperature Temperatures were set at 26℃, 24℃, and 22℃ during the initial, middle, and hardening-off stages of afforestation, respectively, with daytime temperatures adjusted accordingly. Take at night , Take 2℃. The upper limit of the target temperature is set as follows: Determined, where C represents the indoor carbon dioxide concentration after fusion. According to the concentration range set in the human-machine interface, in this embodiment, when C is below 400 μmol / mol, 26°C is used; when C is between 400 μmol / mol and 700 μmol / mol, 28°C is used; and when C is above 700 μmol / mol, 30°C is used. Normalized value of illumination, saturated light intensity The levels were 300 W / m², 400 W / m², and 500 W / m², respectively, at the three reproductive stages. The corresponding brightness variation coefficient (ILR) in the human-machine interface is set to 4°C; the cumulative light quantity correction is calculated as follows: Calculate and limit the amplitude between -1 and 0, and set the light intensity. The values ​​were 4 MJ / m², 6 MJ / m², and 8 MJ / m², respectively. Take 2℃. If the calculated upper limit is lower than the lower limit, then take the upper limit equal to the lower limit, allowing the temperature range to degenerate into a single target value. (Based on mid-term greening...) C is 520 μmol / mol. 260W / m² Taking the working condition of 2.1 MJ / m² as an example, , , , That is, if the indoor carbon dioxide concentration drops to 350 μmol / mol at any given moment, the upper limit is reduced to 26.0℃; however, if the cumulative light intensity has reached the set value, the upper limit is increased to 28.0℃. The environmental controller 5 does not request heating immediately when the indoor temperature falls below a certain fixed value; instead, it first determines the allowable temperature range. This information is then used to determine whether action is necessary. To prevent sudden target jumps caused by rapid weather changes, the target temperature is also constrained by its rate of change. , The corresponding temperature change rate parameter in the human-machine interface is set to 0.05℃ / min in this embodiment, meaning the target temperature can move a maximum of 3℃ per hour; when cloud cover causes... When the temperature drops from 1 to 0.2 within 1 minute, the upper limit of the target temperature also moves downward by only 0.05℃ / min, thus avoiding the need for a large opening for ventilation to be converted into heating in an instant.

[0048] In step S4, the environmental controller 5 converts the relative humidity, saturated water vapor pressure difference, and air moisture content expressed in g / m³ into the actual water vapor partial pressure of the air. Then participate in the calculation. Saturated vapor pressure is calculated according to... Calculations are made in kPa; actual water vapor partial pressure is calculated based on... Calculations are performed, where RH is represented by 0 to 1; saturated vapor pressure difference is calculated as follows: Calculation; absolute moisture content is based on Conversion, unit is g / m³; dew point temperature according to Reverse calculation. The environmental controller 5 internally stores only the uniform target water vapor state. It is determined by the target temperature and the target saturated water vapor pressure difference for that growth stage, i.e. The target saturated vapor pressure difference is linearly interpolated within a set range based on the normalized value of light intensity. The range is 0.4 kPa to 0.8 kPa in the early stage of greening, 0.5 kPa to 1.0 kPa in the middle stage, and 0.8 kPa to 1.2 kPa in the hardening-off stage. When the indoor temperature is below 18℃, during the nighttime period after sunset, or when the indoor temperature change rate is greater than 0.1℃ / min, the environmental controller 5 selects relative humidity as the humidity control coordinate. When it is during the daytime period and the indoor light intensity is greater than 150 W / m², and seedling transpiration becomes the main influencing factor, the saturated vapor pressure difference is selected as the humidity control coordinate. To prevent switching back and forth near critical conditions, both criteria are set with a minimum holding time of 10 minutes and a corresponding hysteresis. The moisture target is not regenerated at the moment of control coordinate switching; instead, it is determined from the current... Calculated based on the current indoor temperature or With an indoor temperature of 24℃, Taking 2.0 kPa as an example, kPa, before switching kPa is used as the control target, and after switching... As control targets, both correspond to the same air moisture state; the skylight 7, side window 8, and circulating fan 11 do not experience a sudden change in opening degree during switching. The back-calculation results are based on... Limited to 40% to 90%, The amplitude is limited to 0.2 kPa to 1.5 kPa; if the amplitude exceeds the limit, the amplitude value is used to calculate the amplitude. It is updated synchronously to ensure that internal objectives remain consistent. Environmental Controller 5 simultaneously calculates the condensation risk margin. When M is below 2℃, directly use To cover the above objectives, the weighting factor of air moisture component in the scheduling cost is increased from 0.8 to 1.5, giving priority to dehumidification and ventilation requests.

[0049] In step S5, the environmental controller 5 calculates the environmental deviation vector. X is composed of indoor temperature, actual water vapor partial pressure in the air, and indoor carbon dioxide concentration. The temperature component is set to 0 when the indoor temperature falls within the allowable temperature range; below the lower limit, it is the difference between the lower limit and the measured value; above the upper limit, it is the difference between the upper limit and the measured value. This forms a bias definition with a dead zone, avoiding unnecessary actions within the target range. A dead zone of ±0.15 kPa is set for the air water vapor component. The carbon dioxide component has an allowable range of 400 μmol / mol to 800 μmol / mol; exceeding this range, the difference between the nearest endpoints is used. To allow comparison of components with different dimensions within the same cost function, each component is divided by its normalized scale: temperature is set to 1 °C, air water vapor to 0.1 kPa, and carbon dioxide to 100 μmol / mol.

[0050] In step S6, the environmental controller 5 establishes a multi-dimensional environmental action vector for each actuator. Each component represents the rate of change in temperature, air moisture state, carbon dioxide concentration, and light environment caused by the actuator under a unit action quantity, with dimensions of ℃ / min, kPa / min, μmol / (mol·min), and W / (m²·min), respectively. The unit action quantity is taken as 10% opening for skylight 7 and side window 8, one power level for heating mechanism 9, one start for circulating fan 11 and exhaust equipment 12, and a 60s water filling time for automatic water filling mechanism 10. The temperature component of heating mechanism 9 is positive, with 0.15℃ / min at the low power level and 0.30℃ / min at the high power level. Its air moisture component is 0 in direct action, but the temperature increase causes… The increase in relative humidity or saturated vapor pressure difference still has an indirect effect, which is determined by the environmental controller 5 in the prediction stage. The changes are uniformly converted. The air moisture component of the automatic irrigation mechanism 10 is taken as positive 0.05 kPa / min, and the temperature component as negative 0.02 ℃ / min. The absolute values ​​of both the temperature and air moisture components of the circulating fan 11 are relatively small; its main function is to reduce the spatial temperature and humidity differences within the seedling greening area 2. Therefore, the improvement in spatial consistency is separately included in the cost function. The temperature component of the exhaust equipment 12 is calculated as follows... The calculation shows that -0.05℃ / min is its inherent heat dissipation effect. For skylight 7 and side windows 8, the environmental controller 5 does not explicitly designate them as cooling devices, but determines their direction and intensity online based on the indoor-outdoor gradient. , , The coefficient of sunroof 7 , , The initial values ​​were 0.004, 0.003, and 0.05, respectively, and the values ​​for side windows 8 were 0.003, 0.0025, and 0.04, respectively. Furthermore, when the wind speed was greater than 3 m / s, the values ​​increased linearly to 1.5 times the wind speed. Therefore, the same window opening action was identified as having two opposite effects in the early morning when the outdoor temperature was lower than the indoor temperature and in the afternoon when the outdoor temperature was higher than the indoor temperature.

[0051] Whenever an actuator completes a recognizable action and its response delay time has elapsed Then, the environmental controller 5 calculates the change in environmental state during that period. ,because yes Within a time period The total change accumulated under the action of each unit of motion, and This represents the rate of change per unit of motion, therefore the identification value used in this study is... , and according to Update action coefficient, forgetting coefficient Take 0.2, for skylight 7 and side window 8 The following timeframes are used: 5 minutes for the heating mechanism 9, 3 minutes for the automatic water filling mechanism 10, and 3 minutes for the exhaust system 12. The recognizable thresholds are set as follows: the opening degree change of the skylight 7 or side window 8 is no less than 20%; the power level change of the heating mechanism 9 is no less than 1 level; and the water filling time of the automatic water filling mechanism 10 is no less than 30 seconds. Furthermore, the reliability of the sensor nodes participating in the calculation within the recognition window must be no less than 0.6, and no other actuators must change their power levels; otherwise, no update will be performed. Taking the skylight 7 opening from 0% to 60% and the indoor-outdoor temperature difference being -6℃ as an example... The motion is 6 units, and the initial value of the temperature component is... ℃ / min, based on this, predict the temperature change within 5 minutes as follows: If the measured change is -0.48℃, then the identification value for this test is... ℃ / min, the updated temperature component is The temperature / min reading is used by the environmental controller 5 to progressively adjust the "how much temperature and humidity can be changed by opening the skylight 7 once". To prevent model divergence caused by abnormal identification, the amplitude of each action coefficient is limited to 0.3 to 3 times its initial value, and the sign is always determined by the current indoor and outdoor gradient. This process is a deterministic online action coefficient identification, which does not require neural network training. The computational and storage resources required can be handled by existing industrial controllers or embedded controllers. The updated coefficients are written to the power-off retention storage area. After power-on, the saved values ​​are used as the initial values. On the first power-on, the factory default values ​​are used. The current action coefficients of each actuator can be viewed on the human-machine interface.

[0052] In step S7, the environmental controller 5 discretizes each actuator into several action levels. In this embodiment, the skylight 7 and side window 8 each have four opening levels: 0%, 30%, 60%, and 100%. The heating mechanism 9 has three levels: closed, low power, and high power. The circulating fan 11 and exhaust equipment 12 each have two levels: closed and open. The automatic water filling mechanism 10 provides the water filling duration level according to the task request and checks it separately. This forms a total of 192 candidate action combinations (4×4×3×2×2). Combinations that violate equipment safety conditions are then eliminated. Safety conditions include: the skylight 7 must be closed when rainfall is active; the skylight 7 opening must not exceed 30% when the outdoor wind speed is greater than 8 m / s and must be closed when the wind speed is greater than 12 m / s; the side window 8 opening must not exceed 30% when the outdoor temperature is below 5℃; and the heating mechanism 9 and exhaust equipment 12 must not operate at their high levels simultaneously. The number of candidates after elimination is usually less than 100, and the time required to go through them one by one is no more than 50 ms. For each candidate action combination U, press Predict the environment state for the next scheduling cycle and calculate the scheduling cost. ,in To determine the normalized predicted environmental bias after the action, W is a diagonal weight matrix. The weights for temperature, air moisture, and carbon dioxide components are set to 1.0, 0.8, and 0.3, respectively. When the condensation risk margin is less than 2°C, the weight of the air moisture component is increased to 1.5; resource usage cost. The heating mechanism 9 is calculated as follows: high power 1.0, low power 0.5; exhaust equipment 12 is calculated as 0.3; circulating fan 11 is calculated as 0.2; automatic water filling mechanism 10 is calculated based on water filling time; skylight 7 and side window 8 are calculated at 0.005 per 10% opening degree; start-up and shutdown costs. The cost is calculated by accumulating the number of gear changes for candidate combinations relative to the previous scheduling cycle, with each gear change accounting for 0.3. For actuators that have already acted in the previous scheduling cycle but have not reached the minimum hold time of 3 minutes, the cost corresponding to their gear change is calculated at 3 times the cost; conflict cost. A weighting factor of 1.0 is applied when a candidate combination simultaneously includes high-level actions of the actuators at both ends of the conflicting edge. , , We took values ​​of 0.2, 0.3, and 0.5 respectively. The final value was... As the execution command for this scheduling cycle, it is sent to each actuator via the output terminal of the environmental controller 5. The target opening degree of the sunroof 7 and the side window 8 is executed in a closed loop by the opening degree feedback potentiometer.

[0053] In addition to the aforementioned action matrix, the environmental controller 5 establishes an actuator conflict diagram. Using each actuator as a node, conflict edges are dynamically determined based on the current action coefficient in each scheduling cycle. A conflict edge is established between two actuators when their action fingerprints have opposite signs in the direction of the principal component of the environmental deviation vector, and the absolute value of their action intensity in that direction is greater than 0.2 times the required change in that direction. For example, when the principal component of the environmental deviation vector is temperature and heating is required, if the temperature component of heating mechanism 9 is positive while the temperature components of skylight 7 and side window 8 are negative, then the high-power setting of heating mechanism 9 and the 60% and 100% opening settings of skylight 7 and side window 8 are marked as conflict actions. Candidates containing this combination are assigned a conflict cost and are directly eliminated when the conflict intensity exceeds a set value. This conflict is not permanently prohibited. When the environmental deviation vector simultaneously represents the need to increase temperature and decrease air moisture, i.e., a low-temperature, high-humidity condition with a risk of condensation, the environmental controller 5 releases the conflict edge between the heating mechanism 9 and the skylight 7 or side window 8, allowing the high-power setting of the heating mechanism 9 to work in conjunction with the 30% opening setting of the skylight 7 or side window 8. This is because the combined effect of the two is to increase the temperature while expelling excess moisture. Therefore, the linkage between the actuators depends on the current direction of the environmental deviation and the real-time value of the actuator's environmental action fingerprint, rather than pre-written fixed interlocking rules.

[0054] Let's take the actual working conditions of a certain morning as an example to illustrate the complete operation of the above closed loop. At this time, the rice seedlings are in the middle of their greening stage. Indoor temperature 15.5℃, indoor relative humidity 96%, indoor carbon dioxide concentration 680μmol / mol, indoor light intensity 30W / m², cumulative effective light dose 0.05MJ / m², outdoor temperature 8℃, outdoor relative humidity 85%, outdoor light intensity 40W / m², no rainfall, wind speed 1.5m / s. Environmental controller 5 consists of... kPa kPa, , Since the temperature is below the 2℃ threshold, the target water vapor state is taken as... The air moisture content deviation is -0.144 kPa, indicating that dehumidification is required; at the same time , , The calculated upper limit is lower than the lower limit, therefore the allowable temperature range degrades to 24.3℃, with a temperature component deviation of +8.8℃, indicating that heating is required. Since the outdoor temperature is lower than the indoor temperature at this time, the temperature component of skylight 7 is... ℃ / min; the actual outdoor water vapor partial pressure is 0.912 kPa, which is lower than that indoors, therefore its air water vapor content is... Both are measured in units of 10% opening. If only the high-power setting of heating mechanism 9 is selected, the temperature deviation decreases by 0.30℃ per minute, but the air moisture content does not decrease, and as the indoor temperature rises, evaporation on the seedling tray surface will cause the actual moisture partial pressure to continue to rise. If the 100% opening of skylight 7 is selected, the temperature component effect is -0.30℃ / min, which completely cancels out the effect of heating mechanism 9, and the temperature deviation increases instead of decreasing. Its scheduling cost is significantly higher in the predicted environmental deviation item and is therefore excluded. Since there are simultaneous heating and dehumidification needs at this time, the conflict is resolved. The final selected combination is the high-power setting of heating mechanism 9, the 30% opening of skylight 7, and the opening of circulating fan 11. The predicted temperature change is positive 0.21℃ / min, and the air moisture partial pressure change is negative 0.007kPa / min, with the deviations in both directions decreasing simultaneously. This combination remains in place for several subsequent scheduling cycles until the temperature deviation enters the dead zone. During this period, due to the minimum holding time constraint, the skylight 7 will not oscillate between 30% and 0%. After approximately 50 minutes, when the indoor temperature rises to 23.8℃ and the condensation risk margin recovers to 3.5℃, the weight of the air moisture component returns to 0.8, the heating mechanism 9 is reduced to a low power setting, and the skylight 7 is closed.

[0055] Taking the work situation at noon on the same day as another example, at this time Indoor temperature 31℃, outdoor temperature 27℃, indoor carbon dioxide concentration 380μmol / mol, indoor light intensity 520W / m², cumulative effective light intensity 8MJ / m². Environmental controller 5 calculates... Because the carbon dioxide concentration is below 400 μmol / mol, it is taken Therefore The temperature component deviation is -5.0℃. The temperature component of skylight 7 is... ℃ / min, side window 8 is -0.012℃ / min, exhaust equipment 12 is The heating mechanism 9 was eliminated because its direction of action was opposite to the demand. The final selected combination was a 100% opening of skylight 7, 100% opening of side windows 8, and the ventilation system 12, with a predicted temperature change of -0.45℃ / min. Meanwhile, the outdoor carbon dioxide concentration was 420 μmol / mol, higher than the indoor concentration. Ventilation caused the indoor carbon dioxide concentration to rise again; once it exceeded 400 μmol / mol... As the temperature rises to 28.0℃, the permissible temperature range widens, the temperature deviation decreases, and the ventilation demand decreases accordingly, thus avoiding the need to maintain high ventilation levels for extended periods under low carbon dioxide concentration conditions.

[0056] For management commands and internal linkage tasks issued by the human-machine interface, the environmental controller 5 does not directly output them to the relays, but first converts them into task requests. The process is then arbitrated by the candidate combination construction step in step S7. Taking an automatic irrigation task as an example, the task request type is irrigation, priority is 3, duration is 120s, and the latest completion time is 09:30. Before generating the irrigation control quantity, the environmental controller 5 sequentially checks whether there is an equipment safety alarm, whether the equipment is in a test state where irrigation is prohibited, whether the condensation risk margin is lower than 2℃, and whether there is a higher priority safety ventilation or equipment protection task. In the above-mentioned early morning conditions, because the condensation risk margin is only 0.6℃, the task is postponed until the margin rises to 3.5℃ before outputting the control quantity to the automatic irrigation mechanism 10. During irrigation, the circulating fan 11 is turned on simultaneously to accelerate the local water vapor diffusion. If the conditions are not met by the latest completion time of 09:30, the minimum irrigation duration of 60s is set, and the opening of the skylight 7 is increased by one level to ensure that the process requirement of water supply to the seedling trays is not postponed indefinitely. Therefore, human-machine interface operation, automatic time tasks, and internal linkage tasks all obtain execution authority through the same scheduling link, avoiding the three sets of control logics from competing for the same execution mechanism at the same time.

[0057] When the sensor status is abnormal, the environmental controller 5 operates in a degraded manner according to the three-level mechanism. Under normal control conditions, all scheduling components participate in the calculation. When fewer than two reliable nodes among the indoor temperature and humidity nodes cause the air moisture status to fail, it enters a restricted control state, ceasing fine humidity scheduling based on saturated vapor pressure difference, retaining only safety control based on temperature, and limiting the maximum opening of skylight 7 and side windows 8 to 60%. When the indoor carbon dioxide concentration fails, the highest permissible temperature corresponding to the higher concentration range is prohibited, i.e. The lowest setting of 26℃ is always used to avoid maintaining an excessively high target temperature when the concentration is unknown. When the light channel of the outdoor environment sensor unit 4 fails, the average of the sunrise and sunset times of the most recent 5 days is used to replace the measured value of the day to generate the photoperiod phase. The cumulative effective light quantity is estimated based on the average of the same period, and the effect coefficient of the cumulative light quantity correction is halved. When all indoor temperature signals fail, that is, when the sum of the reliability of the physical quantity is lower than the set lower limit, the system enters a safety control state, shuts down the heating mechanism 9 and the automatic water filling mechanism 10, the skylight 7 and side window 8 operate at a preset 30% safety opening, the circulating fan 11 continues to run, and an alarm is output through the communication unit 6, maintaining the above state until manual confirmation. The entry and exit of the degraded state are both set with a confirmation time of not less than 10 minutes to avoid frequent switching of control state due to instantaneous communication jitter.

[0058] In other embodiments, the environmental actuator can be equipped with a sunshade, a wet curtain fan, or a supplementary lighting device. It only needs to establish a corresponding actuator environmental action fingerprint to be included in the same scheduling loop. The sunshade has a negative light environment component and a negative temperature component during periods of strong light, while the supplementary lighting device has a positive light environment component and a positive temperature component. The saturated vapor pressure can also be calculated using other empirical formulas, as long as it remains consistent within the same environmental controller 5. When the number of actuators is large, significantly increasing the candidate combination size, the actuators with the highest action intensity can be selected first based on the principal component of the environmental deviation, and then traversed. Alternatively, an offline lookup table can be used instead of online traversal, or a greedy optimization method can be used for each actuator, followed by back-substitution verification. All these methods do not change the processing order of first establishing the environmental deviation and then calculating the device combination. In addition to RS485 bus, the communication unit 6 can also use Ethernet or wireless methods, and the number of sensor nodes can be expanded between 2 and 16. The scheduling period Δt can be adjusted between 30s and 300s. The heating mechanism 9, with a slower response, can use an action period that is an integer multiple of the scheduling period. The seedling growth stage variable G can be input by the operator or automatically calculated from sowing batch information or the tray entry count of the seedling tray conveyor line. None of the above modifications change the control structure: first, a dynamic environmental target domain is generated from the environmental state; then, the actuator's environmental action fingerprint is used to select the equipment combination, and finally, conflict arbitration is performed before outputting the result.

[0059] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Equivalent substitutions or improvements made by those skilled in the art without departing from the concept of the present invention should be included within the protection scope of the present invention.

Claims

1. An environmentally controlled seedling greening greenhouse, comprising a greenhouse body, a seedling greening area set within the greenhouse body, an indoor environmental sensing unit, an outdoor environmental sensing unit, a communication unit, an environmental actuator, and an environmental controller connected to the indoor and outdoor environmental sensing units via the communication unit and connected to the environmental actuator, wherein the environmental actuator includes at least two of the following: a skylight, a side window, a heating mechanism, and an automatic irrigation mechanism, characterized in that... The environment controller is configured to perform the following operations according to a scheduling cycle: For similar sensing nodes, the consistency of their time change rate, spatial deviation, and consistency of physical quantities are calculated respectively. Based on this, the reliability of each sensing node is determined. The similar detection data are then weighted and fused according to the reliability to obtain environmental state quantities including indoor temperature, indoor air moisture state, indoor carbon dioxide concentration, indoor light intensity, outdoor temperature, outdoor air moisture state, and outdoor light intensity. The sunrise and sunset times for the day are determined based on the outdoor light intensity, a preset light threshold, and the duration of the light condition. The results for the day are constrained by the average of the most recent N days. The daytime is normalized using the sunrise and sunset times as endpoints to obtain the photoperiod phase. The cumulative effective light intensity for the day is obtained by integrating the outdoor light intensity from the sunrise time to the current time. The preset time adjustments before sunrise, after sunrise, before sunset, and after sunset are used as offsets relative to the sunrise and sunset times. The dynamic environmental target domain is generated by using the seedling growth stage, the photoperiod phase, the cumulative effective light intensity of the day, and the indoor carbon dioxide concentration as scheduling coordinates. The dynamic environmental target domain includes the allowable temperature range defined by the lower limit of the target temperature and the upper limit of the target temperature, as well as the target water vapor state expressed by the actual water vapor partial pressure of the air. The indoor temperature and the indoor air moisture state are uniformly converted into the actual water vapor partial pressure of the air as a common intermediate variable for humidity control. Based on the environmental state quantity and the dynamic environmental target domain, calculate an environmental deviation vector containing temperature, air moisture, and carbon dioxide components. The actuator environmental action fingerprint of each actuator is stored. The actuator environmental action fingerprint characterizes the direction and intensity of the actuator’s effect on temperature, air moisture state and carbon dioxide concentration per unit amount of action. The action direction of the skylight and side window is determined online based on the difference between indoor and outdoor temperature and the difference between indoor and outdoor air moisture state. Candidate action combinations are constructed based on the discrete action levels of each actuator. The environmental state after execution of each candidate action combination is predicted based on the actuator environmental action fingerprint, and the scheduling cost is calculated. After eliminating candidate action combinations whose main action directions cancel each other out, the one with the lowest scheduling cost is selected and output to the corresponding actuator. After the actuator completes the action and the response delay time of the actuator has elapsed, the actuator environmental action fingerprint is updated with weight by the ratio of the change in the environmental state quantity to the product of the corresponding action quantity and the response delay time.

2. The environmentally controlled seedling greenhouse according to claim 1, characterized in that: The indoor environment sensing unit includes multiple sensing nodes, which are respectively arranged at different locations in the seedling greening area and near the skylight and the air outlet of the heating mechanism, with the installation height located near the seedling canopy. The environment controller is also configured to reduce the weight of a sensing node in the weighted fusion when the confidence level of the sensing node is lower than a first threshold, remove its detection data and output an alarm when the confidence level of the sensing node is lower than a second threshold, maintain the environmental state quantity of the previous scheduling cycle and set a degradation flag when the sum of the confidence levels of similar sensing nodes is lower than a set lower limit, stop humidity scheduling based on saturated water vapor pressure difference and only retain temperature safety control when the indoor air moisture state fails, prohibit the use of the highest allowable temperature corresponding to the higher carbon dioxide concentration range when the indoor carbon dioxide concentration fails, and generate the photoperiodic phase by replacing the measured value of the day with the average of the sunrise and sunset event times of the most recent N days when the outdoor light intensity fails.

3. The environmentally controlled seedling greening greenhouse according to claim 1, characterized in that: The environmental actuators also include a circulating fan installed above the seedling greening area and an exhaust fan installed on the main gable wall of the greenhouse. The skylights and side windows are respectively equipped with opening feedback elements. The discrete action positions of each actuator include a closed position and at least one active position. The environmental controller is also configured to limit the opening position of the skylights based on the outdoor rainfall and outdoor wind speed, and to eliminate candidate action combinations that violate equipment safety conditions. When calculating the scheduling cost, for actuators that have been activated in the previous scheduling cycle but have not reached the minimum hold time, the start-stop cost item corresponding to the change of their position is increased.

4. A multi-dimensional parameter scheduling method for an environmentally controlled seedling greening greenhouse as described in claim 1, characterized in that, Includes the following steps: S1. The environmental controller collects the detection data of the indoor and outdoor environmental sensing units according to the scheduling cycle. For similar sensing nodes, the consistency of their time change rate, spatial deviation, and physical quantity consistency are calculated respectively. Based on this, the reliability of each sensing node is determined. The similar detection data are weighted and fused according to the reliability to obtain environmental state quantities including indoor temperature, indoor air moisture state, indoor carbon dioxide concentration, indoor light intensity, outdoor temperature, outdoor air moisture state, and outdoor light intensity. S2. Determine the sunrise and sunset times for the day based on the outdoor light intensity, the preset light threshold, and the duration of the light condition, and apply a deviation constraint to the determination results for the day using the average of the sunrise and sunset times of the most recent N days. Using the sunrise and sunset times as endpoints, the daytime is normalized to obtain the photoperiod phase representing the position of the current time in the photoperiod of the day. The outdoor light intensity from the sunrise time to the current time is integrated to obtain the cumulative effective light intensity of the day. The preset time adjustments before sunrise, after sunrise, before sunset, and after sunset are used as offsets relative to the sunrise and sunset times to form the device control range that drifts with the daytime light environment. S3. A dynamic environmental target domain is generated using the seedling growth stage, the photoperiod phase, the cumulative effective light intensity of the day, and the indoor carbon dioxide concentration as scheduling coordinates. The lower limit of the target temperature is determined by the reference temperature corresponding to the seedling growth stage and the correction amount of the photoperiod phase. The upper limit of the target temperature is the smaller of the highest allowable temperature corresponding to the carbon dioxide concentration range and the sum of the lower limit of the target temperature, the normalized value of the light intensity, and the correction amount of the cumulative effective light intensity. The change in target temperature between adjacent scheduling cycles is limited by a preset temperature change rate. S4. Convert the indoor temperature and indoor air moisture state into the actual water vapor pressure of the air, and use the actual water vapor pressure of the air as the common intermediate variable for humidity control; when switching between relative humidity as the humidity control coordinate and saturated water vapor pressure difference as the humidity control coordinate, keep the target water vapor state unchanged, and calculate the target humidity control value after switching based on the saturated water vapor pressure corresponding to the current indoor temperature from the switched humidity control coordinate. S5. Calculate an environmental deviation vector containing temperature, air moisture, and carbon dioxide components based on the environmental state quantity and the dynamic environmental target domain. S6. Establish an actuator environmental action fingerprint for each actuator. The actuator environmental action fingerprint characterizes the direction and intensity of the actuator's action on temperature, air moisture state and carbon dioxide concentration under a unit amount of action. The action direction of the skylight and side window is determined online based on the difference between indoor and outdoor temperature and the difference between indoor and outdoor air moisture state. S7. Construct candidate action combinations according to the discrete action levels of each actuator, predict the environmental state after each candidate action combination is executed based on the actuator environmental action fingerprint, and calculate the scheduling cost including the predicted environmental deviation item, resource usage cost item, start-stop cost item and conflict cost item. After eliminating candidate action combinations whose main action directions cancel each other out, select the candidate action combination with the smallest scheduling cost as the execution instruction output for this scheduling cycle. S8. After the actuator completes the action and the response delay time of the actuator has elapsed, the change in the environmental state quantity is collected. The actuator environmental action fingerprint is updated by weighting the ratio of the change quantity to the product of the corresponding action quantity and the response delay time, and then the process returns to step S1.

5. The multi-dimensional parameter scheduling method for an environmentally controlled seedling greening greenhouse according to claim 4, characterized in that: In step S2, when the outdoor light intensity exceeds the preset sunrise light threshold and the duration reaches the first set duration, the moment of exceeding the threshold is determined as the sunrise event moment of the day. When the outdoor light intensity falls below the preset sunset light threshold and the duration reaches the second set duration, the moment of falling below the threshold is determined as the sunset event moment of the day. When the determined time deviates from the corresponding average of the most recent N days beyond the allowable range, the sunrise or sunset time of the day is not updated, and the photoperiod phase is determined by the average and the current duration of continuous illumination.

6. The multi-dimensional parameter scheduling method for an environmentally controlled seedling greening greenhouse according to claim 4, characterized in that: In step S3, the correction amount of the optical period phase increases as the optical period phase progresses from sunrise to noon and decreases as it progresses from noon to sunset; The cumulative effective light quantity correction value is taken as a non-positive value when the cumulative effective light quantity on the day is lower than the set light quantity corresponding to the seedling growth stage; The maximum permissible temperature corresponding to a lower indoor carbon dioxide concentration range is lower than the maximum permissible temperature corresponding to a higher indoor carbon dioxide concentration range. When the upper limit of the target temperature is lower than the lower limit of the target temperature, the upper limit of the target temperature shall be equal to the lower limit of the target temperature.

7. The multi-dimensional parameter scheduling method for an environmentally controlled seedling greening greenhouse according to claim 4, characterized in that: In step S4, when the indoor temperature is lower than the set temperature, it is in the nighttime interval after the sunset event, or the rate of change of indoor temperature is greater than the set value, relative humidity is selected as the humidity control coordinate. When it is daytime after sunrise and the indoor light intensity is greater than the set light value, the saturated water vapor pressure difference is selected as the humidity control coordinate. The condensation risk margin is determined based on the difference between the indoor temperature and the dew point temperature determined by the actual water vapor partial pressure of the air. When the condensation risk margin is less than the set threshold, the weighting coefficient of the air water vapor component in the scheduling cost of step S7 is increased.

8. The multi-dimensional parameter scheduling method for an environmentally controlled seedling greening greenhouse according to claim 4, characterized in that: In step S6, the sign of the temperature component of the actuator environmental fingerprint of the sunroof and side window is consistent with the sign of the difference between the outdoor temperature and the indoor temperature, and the sign of the air moisture component is consistent with the sign of the difference between the outdoor air moisture state and the indoor air moisture state. Moreover, the intensity of the effect increases with the increase of the absolute value of the corresponding difference and the opening degree. The temperature component in the environmental action fingerprint of the actuator of the heating mechanism is positive, and its air water vapor component is determined according to the changes in relative humidity and saturated water vapor pressure difference caused by the change in saturated water vapor pressure after the indoor temperature rises. The air moisture component in the fingerprint of the actuator of the automatic watering mechanism is positive. In step S8, the weighted update is performed only when the change in the amount of action is greater than the set identifiable threshold and the confidence level of the sensor node involved in the calculation is not lower than the first threshold. The weighted update performs a first-order weighting on the current identification result and the historical actuator environment action fingerprint according to the forgetting coefficient. The forgetting coefficient is greater than 0 and less than 1, and the updated actuator environment action fingerprint is limited to a set multiple of its initial value.

9. The multi-dimensional parameter scheduling method for an environmentally controlled seedling greening greenhouse according to claim 4, characterized in that: In step S7, an actuator conflict diagram is established with each actuator as a node. When the actuator environmental action fingerprints of two actuators have opposite directions of action in the principal component direction of the environmental deviation vector, and the intensity of their actions in that direction is greater than the set proportion of the required change in that direction, a conflict edge is established between them, and candidate action combinations that simultaneously include the high-level actions of the actuators at both ends of the conflict edge are eliminated. When the environmental deviation vector simultaneously represents the need to increase the temperature and decrease the air moisture state, the conflict edge between the heating mechanism and the sunroof or side window is released, allowing the action of the heating mechanism to be combined with the low opening level of the sunroof or side window.

10. The multi-dimensional parameter scheduling method for an environmentally controlled seedling greening greenhouse according to claim 4, characterized in that: It also includes converting external management instructions and internal linkage tasks into task requests that include task type, priority, duration and latest completion time, and adding the task requests as constraints to the candidate action combination construction in step S7. When there is a device safety alarm, the device is in a testing state or the condensation risk margin is less than a set threshold, the task request corresponding to the automatic watering mechanism is postponed until the constraint is lifted or the latest completion time is reached.

Images