Malus spectabilis environment regulation and health monitoring method based on edge internet of things

Abstract

The invention relates to the technical field of intelligent flower cultivation, and discloses a begonia flower environment regulation and health monitoring method based on edge internet of things. According to the method, a dual-drive model in which an edge internet-of-things perception model and a begonia health assessment model are coordinated is constructed, environmental data such as temperature, humidity, illumination and soil nutrients are collected in real time through the edge internet-of-things perception model, and the growth state and health risk of begonia are accurately recognized by means of the health assessment model. An edge end intelligent regulation and control module is designed based on a dual-drive model, and localized linkage of monitoring and regulation and control is achieved. Layered progressive environment regulation and control are executed according to an optimization target, and environment equipment is adjusted in a differentiated mode according to specific requirements of different areas and different plants. According to the method, the delay and bandwidth problems of traditional cloud transmission are avoided, the regulation response speed and accuracy are improved, resource waste is reduced, the occurrence probability of diseases and pests is reduced, and the growth quality and ornamental value of the begonia flowers are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent flower cultivation, in particular to a method for environmental regulation and health monitoring of Chinese flowering crabapple based on edge Internet of Things. BACKGROUND

[0002] As a traditional ornamental flower in China, Chinese flowering crabapple is widely used in landscape and home gardening due to its ornamental value and cultural implications. However, the growth of Chinese flowering crabapple is sensitive to environmental conditions, and slight changes in temperature, humidity, light intensity, and soil fertility can affect its growth status and even cause diseases and pests. Traditional cultivation and management of Chinese flowering crabapple relies on manual experience, and growers judge the growth status by observing plant morphology and leaf color, and then adjust irrigation, fertilization, and temperature control accordingly. This method is not only time-consuming and labor-intensive, but also subjective, making it difficult to accurately grasp the timing and magnitude of environmental regulation, often leading to resource waste or poor growth. With the development of agricultural intelligence, some regions have begun to use Internet of Things technology for flower environment monitoring, collecting environmental data through sensors and transmitting them to the cloud for analysis and processing. However, such systems often have high data transmission delays and rely on cloud computing resources. When the monitoring area is large or the number of sensors is large, uploading massive amounts of data will consume a lot of network bandwidth, and the response speed of cloud processing cannot meet the real-time regulation requirements. In addition, existing monitoring models focus on monitoring a single environmental factor, lack depth analysis of the correlation between Chinese flowering crabapple growth status and environmental factors, and cannot accurately assess plant health, affecting the effectiveness of regulation measures. Current environmental regulation equipment uses a unified control strategy, which cannot differentiate between the specific needs of individual or small-area Chinese flowering crabapple. For example, Chinese flowering crabapple at different locations in the same greenhouse may exhibit different growth states due to uneven light, but the regulation system still performs the same light regulation scheme, resulting in excess light for some plants and insufficient light for others. This "one-size-fits-all" regulation mode not only fails to fully adapt to the growth characteristics of Chinese flowering crabapple, but also causes unnecessary consumption of energy and resources.

[0003] In existing technologies, environmental monitoring and health assessment are often independent of each other, and there is a lack of real-time linkage mechanism between monitoring data and assessment results. When the sensor detects an environmental anomaly, manual judgment or complex procedures are required to trigger regulation operations, and during this process, the environment may have already caused irreversible damage to the Chinese flowering crabapple. Especially in the prevention and control of diseases and pests, traditional methods often take measures after the disease appears, at which point the plant has already been damaged, and the prevention and control effect is greatly reduced. Therefore, how to achieve real-time monitoring of environmental factors, accurate assessment of plant health, and rapid response of regulation measures has become a problem to be solved in the cultivation and management of Chinese flowering crabapple. SUMMARY

[0004] The present application aims to provide a method for environmental regulation and health monitoring of Malus hupehensis based on edge Internet of Things, to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides a method for environmental regulation and health monitoring of Malus hupehensis based on edge Internet of Things, which comprises: S1: constructing a double-driven model cooperated by an edge Internet of Things perception model and a Malus hupehensis health assessment model; S2: monitoring the growth process of Malus hupehensis by using the double-driven model, and designing an edge intelligent regulation module according to the double-driven model; S3: optimizing the regulation strategy according to the optimization target, performing hierarchical and progressive environmental regulation, and adjusting the environmental regulation equipment.

[0006] Preferably, the construction step of the edge Internet of Things perception model comprises: Step S101: analyzing the growth environment characteristics of Malus hupehensis, and determining the characteristic parameters; The characteristic parameters include environmental characteristic parameters and physiological characteristic parameters; the environmental characteristic parameters include temperature, humidity and light intensity, and the physiological characteristic parameters include leaf color, stem diameter and flower opening degree; Step S102: analyzing the action mechanism of Malus hupehensis environmental regulation and health monitoring, and constructing an edge Internet of Things perception model; The edge Internet of Things perception model is composed of an environment collection model, a data transmission model and a threshold discrimination model; Step S103: simulating the growth process of Malus hupehensis by using a numerical calculation method, and optimizing the edge Internet of Things perception model according to the simulation results; The equation of each model in the edge Internet of Things perception model is converted into a discrete mathematical form; the initial conditions are set; the numerical calculation method is used to solve the discrete equation, to simulate the environmental data evolution process in each stage of the growth process of Malus hupehensis; the simulation results are post-processed; and the model parameters of the edge Internet of Things perception model are optimized according to the difference between the post-processed simulation results and the characteristic parameters.

[0007] Preferably, the construction process of the environment collection model is as follows: determining the action mechanism of the environment collection model, which describes the real-time collection process of environmental parameters; selecting the environment collection model; setting the time, space and frequency balance rules of data collection, simulating the continuous collection process of environmental parameters; and calibrating the environment collection model by corresponding characteristic parameters; The construction process of the data transmission model is: determining the action mechanism of the data transmission model, the action mechanism of the data transmission model is to describe the data interaction process between the edge and the cloud; selecting a data transmission model; establishing delay, packet loss and bandwidth balance rules of data transmission to simulate the stability of data transmission; calibrating the data transmission model through corresponding characteristic parameters. The construction process of the threshold discrimination model is: determining the action mechanism of the threshold discrimination model, the action mechanism of the threshold discrimination model is to describe the correlation discrimination process between environmental parameters and health indicators; selecting a threshold discrimination model; establishing the parameter range and correlation rules corresponding to threshold discrimination; calibrating the threshold discrimination model through corresponding characteristic parameters.

[0008] Preferably, the construction method of the edge intelligent regulation and control module comprises: Defining the state space as the real-time collected growth parameter vector, including environmental parameters and physiological parameters of the Chinese flowering crabapple; Defining the action space as the adjustable parameter set of the environmental regulation and control equipment, including the light supplement duration adjustment amount, the irrigation amount increment and the continuous control amount of the ventilation frequency; Building an Actor-Critic network architecture containing a communication reliability verification gate, wherein the Actor network is used to generate continuous regulation and control actions, and the Critic network is used to design a multi-objective reward function and evaluate the multi-objective reward; The design method of the multi-objective reward function comprises: The suitability improvement term, the health deterioration penalty term and the energy consumption exceeding gradient penalty term are constructed by combining the environmental suitability, the health index and the regulation and control energy consumption through a dynamic weight coefficient, and the multi-objective reward function is obtained according to the suitability improvement term, the health deterioration penalty term and the energy consumption exceeding gradient penalty term, wherein the health deterioration penalty strength nonlinearly increases with the deviation of temperature and humidity from the optimal interval.

[0009] Preferably, the implementation method of the communication reliability verification gate comprises: adding a constraint verification module to the output layer of the Critic network, the constraint verification module performs nonlinear fusion on the predicted Q value and the communication reliability coefficient through a neural network layer, and outputs the action value evaluation result modified by the communication constraint.

[0010] Preferably, the adjustment method of the dynamic weight coefficient comprises: The monitoring data in a preset period is taken as an input of a stage identification model to obtain a current stage, and the current stage includes a budding stage, a flowering stage and a fruiting stage; in the budding stage, a dynamic weight coefficient corresponding to the environmental suitability is increased to a preset weight threshold, and dynamic weight coefficients corresponding to a health degradation penalty term and an energy consumption exceeding gradient penalty term are randomly allocated according to a constraint condition that the sum of all dynamic weight coefficients is 1; in the flowering stage, the dynamic weight coefficients corresponding to the health degradation penalty term and the energy consumption exceeding gradient penalty term are dynamically allocated according to a health index quota; and in the fruiting stage, an emergency adjustment factor is introduced to adjust the dynamic weight coefficient corresponding to the energy consumption exceeding gradient penalty term.

[0011] Preferably, the calibration method of the health degradation penalty term comprises: Gradient experiments are performed in different temperature and humidity intervals of the Chinese flowering crabapple, leaf wilting rates at each interval point are collected, a temperature and humidity influence factor is calibrated based on the leaf wilting rates at each interval point, a mapping relationship between temperature and humidity deviation and health degradation penalty gain coefficient is established, and the health degradation penalty term is calibrated based on the mapping relationship between temperature and humidity deviation and health degradation penalty gain coefficient.

[0012] Preferably, the optimization target of the optimization target optimization control strategy comprises maximizing the environmental parameter suitability, maximizing the Chinese flowering crabapple health index and minimizing the control energy consumption.

[0013] Preferably, the method for performing hierarchical progressive environmental control comprises: Pre-training under the constraint of a collaborative model is performed in a virtual environment to generate an initial control strategy library; Data-driven strategy fine-tuning is achieved through a double experience replay pool; The action space constraint is dynamically relaxed according to a running stability index; The double experience replay pool comprises a data compliance pool and an actual optimization pool; the data compliance pool is used to store experience samples that meet preset constraint conditions; and the actual optimization pool is used to store effective samples in a running process; wherein, a sampling priority of the double experience replay pool is dynamically determined by a weighted fusion result of a data compliance score and a reward value.

[0014] Preferably, the method for dynamically relaxing the action space constraint comprises: When the environmental parameter fluctuation rate of consecutive N control periods is not higher than a preset fluctuation rate threshold, the constraint parameter is updated; and the method for updating the constraint parameter comprises: The action space constraint parameter is adaptively relaxed and adjusted according to a running stability index, and when a performance index reaches a preset index threshold, the allowed control amount change range is proportionally expanded.

[0015] Compared with the prior art, the present application has the beneficial effects that: The method realizes the deep integration of environmental monitoring and health assessment by constructing a double-driven model of edge Internet of Things perception model and health assessment model of Chinese flowering crabapple. The edge Internet of Things perception model can collect multi-dimensional environmental data such as temperature, humidity, light, and soil nutrients in real time locally, avoiding data delay and bandwidth occupation problems in traditional cloud transmission mode, making the acquisition of environmental information more efficient and timely. At the same time, the health assessment model can accurately identify the growth state of the plant based on real-time data collected by the edge, including leaf expansion, flower color saturation, root activity and other key indicators, so as to discover potential health risks such as nutrient imbalance and early signs of disease, breaking the limitations of traditional manual observation. The edge intelligent control module designed based on the double-driven model realizes the localization of monitoring and control. The module can generate targeted control instructions directly according to the health assessment results without relying on cloud processing, greatly improving the control response speed. Unlike traditional unified control strategies, hierarchical and progressive environmental control can adjust the environment equipment in stages and differently according to the specific needs of different regions and plants. For example, for areas with insufficient light, gradually increase the light intensity to the appropriate range while fine-tuning the humidity to cooperate with photosynthesis; for areas with acidic soil, apply the modifier in batches to avoid stimulating the roots with one-time adjustment. This fine-tuned control method not only fully meets the environmental needs of Chinese flowering crabapple at different growth stages, but also reduces energy and resource waste and lowers cultivation and management costs. The method combines edge computing and Internet of Things technology to ensure monitoring and control accuracy while reducing dependence on network environment, allowing the system to operate stably in areas with weak network signals or remote areas. The cooperative working mode of the double-driven model not only improves data processing efficiency but also enhances system adaptability, allowing model parameter adjustment according to different characteristics of Chinese flowering crabapple varieties to adapt to the growth needs of different varieties of Chinese flowering crabapple. Through this closed-loop monitoring-evaluation-control mechanism, the influence of environmental fluctuations on Chinese flowering crabapple growth can be effectively reduced, the probability of disease and pest occurrence can be reduced, and the overall growth quality and ornamental value of the plant can be improved, providing a new technical path for the large-scale and intelligent cultivation of Chinese flowering crabapple. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The working principle diagram of the Chinese flowering crabapple environmental regulation and health monitoring method based on edge Internet of Things described in the present application; Figure 2 The flowchart of the construction steps of the edge Internet of Things perception model; Figure 3 The flowchart of the adjustment method of the dynamic weight coefficient; Figure 4 Flow chart of the method for calibrating the health deterioration penalty term. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0018] Please refer to Figures 1-4 The present application provides a kind of based on edge Internet of Things malus speciosa environmental regulation and health monitoring method, specific implementation steps are as follows: S1: construct the double drive model coordinated by edge Internet of Things perception model and malus speciosa health assessment model.Edge Internet of Things perception model is responsible for collecting and processing the environmental and physiological data of malus speciosa growth, and malus speciosa health assessment model assesses the health status of malus speciosa based on these data, and the two work together to form a mutually coordinated double drive mechanism, providing basic model support for subsequent monitoring and regulation.

[0019] S2: using double drive model to monitor malus speciosa growth process, according to the design of edge intelligent regulation and control module of double drive model. Double drive model continuously tracks the changes of various parameters of malus speciosa growth in the running process, and reflects its growth state and the situation of the environment. Based on the output results and working logic of double drive model, design the regulation and control module that can realize intelligent decision and control at edge, which can directly interface with monitoring data and generate regulation and control instructions according to preset rules and algorithms.

[0020] S3: optimize the regulation and control strategy according to the optimization target, implement hierarchical progressive environmental regulation, and adjust the environmental regulation equipment. Optimization target covers environmental parameter suitability, malus speciosa health index and regulation energy consumption, etc. Through specific algorithm, the regulation and control strategy is optimized, so that the regulation and control measures are more in line with the growth needs of malus speciosa and energy efficient. Then, according to the hierarchical progressive way, implement environmental regulation, gradually adjust environmental parameters, and finally realize the precise adjustment of irrigation equipment, light supplementing equipment, ventilation equipment and other environmental regulation equipment.

[0021] Embodiment 1: Step S101: Analyze the growth environment characteristics of Chinese flowering crabapple and determine the characteristic parameters. The characteristic parameters are divided into two categories: environmental characteristic parameters and physiological characteristic parameters. Environmental characteristic parameters include temperature, humidity, and light intensity. Temperature is directly related to the activity of enzymes in Chinese flowering crabapple, affecting the intensity of photosynthesis and respiration; humidity affects water absorption and nutrient transport by affecting transpiration and stomatal opening and closing; light intensity determines the rate of photosynthesis, and the demand for light varies at different growth stages. Physiological characteristic parameters include leaf color, stem diameter, and flower opening degree. Leaf color can reflect chlorophyll content and nutritional status, and uniform and bright color is related to healthy state; changes in stem diameter reflect growth rate and support capacity, and the stability of diameter growth is related to growth conditions; flower opening degree directly shows the growth process of flowering period, and the change of opening degree reflects the reproductive growth state from bud to full bloom.

[0022] Step S102: Analyze the mechanism of environmental regulation and health monitoring of Chinese flowering crabapple, and build an edge Internet of Things perception model. The model consists of an environment collection model, a data transmission model, and a threshold discrimination model. The environment collection model focuses on real-time capture of environmental parameters such as temperature, humidity, and light intensity, continuously acquiring data according to the set rules through sensor nodes distributed in the planting area, forming a dynamic perception of the environmental state. The data transmission model is responsible for data interaction between the edge and the cloud, handling the flow path, format conversion, and state monitoring of data in the transmission process, ensuring that the collected environmental and physiological data can be accurately and timely transmitted between the two ends. The threshold discrimination model establishes the association logic between environmental parameters and health indicators, compares and analyzes environmental data with physiological parameters such as leaf color and stem diameter through pre-set parameter ranges and judgment rules, and identifies whether Chinese flowering crabapple is in a normal growth state.

[0023] Step S103: Simulate the growth process of the Malus hupehensis using numerical calculation method, and optimize the edge IoT sensing model according to the simulation results. First, convert the equation of each model in the edge IoT sensing model into discrete mathematical form, decompose the continuous time and space variables into a series of discrete nodes, so that the model can be solved by numerical calculation. Set the initial conditions, including the initial values of the environmental parameters at the initial stage of Malus hupehensis planting, such as setting the initial temperature as the average air temperature of the planting area, the initial humidity as the baseline humidity of the soil, and the initial physiological state parameters of Malus hupehensis, such as the stem diameter, leaf color grade and other parameters at the seedling stage. Use numerical calculation method to solve the discrete equation, simulate the evolution process of environmental data such as temperature and humidity over time during different growth stages of Malus hupehensis, such as from seedling to mature plant, and how these environmental changes affect the changes of physiological parameters such as leaf color and stem diameter. Post-process the simulation results, including data organization, outlier removal and trend analysis, to form a complete simulation curve of the growth process. According to the differences between the simulation results after post-processing and the actual monitored characteristic parameters, such as the deviation between the simulated temperature change curve and the actual temperature data collected by the sensor, or the difference between the simulated leaf color change and the actual observed color change, adjust the model parameters of the edge IoT sensing model. If the simulation data of the environmental collection model deviates greatly from the actual collection data, correct the collection frequency coefficient in the model; if the judgment result of the threshold judgment model does not match the actual health status, adjust the judgment threshold parameter in the model, and through multiple iterations of optimization, make the output results of the edge IoT sensing model closer to the actual situation of Malus hupehensis growth, and improve the sensing accuracy of the model for environmental and physiological parameters. This process needs to be repeated for simulation, comparison and correction until the difference between the simulation results of the model and the actual characteristic parameters is within an acceptable range, ensuring that the model can provide reliable basic data support for subsequent monitoring and regulation. During the optimization process, the characteristics of different growth stages need to be considered, and the model parameters need to be adjusted according to the growth characteristics of different stages such as germination stage, flowering stage and fruiting stage, so that the model can maintain good sensing performance at each stage. At the same time, the calculation efficiency of the model also needs to be considered, that is, on the premise of ensuring the simulation accuracy, avoid wasting calculation resources caused by excessive complex parameter settings, and ensure that the model can run efficiently on the edge, meeting the needs of real-time monitoring and regulation.

[0024] In the construction process of the environment collection model, first of all, its action mechanism lies in the complete description of the real-time collection process of environmental parameters. A collection model suitable for the growth environment of Chinese flowering crabapple is selected, which needs to be adapted to multiple sensor types, including temperature sensors, humidity sensors, and light intensity sensors. Time, space, and frequency balance rules for data collection are established. In the time dimension, combined with the diurnal growth rhythm of Chinese flowering crabapple, different collection periods are divided, such as shortening the collection interval during the day when the light is strong, and appropriately extending it at night. In the spatial dimension, according to the size and layout of the planting area, sensors are placed at different positions around the Chinese flowering crabapple plant to ensure coverage of key parts such as plant roots, leaves, and flowers. In the frequency dimension, considering the energy consumption of sensors and the real-time data demand, a reasonable collection frequency is determined to simulate the continuous collection process of environmental parameters. The collected temperature, humidity, light intensity, and other characteristic parameters are used to calibrate the environment collection model. By comparing the collected data with the sensor measured data, the collection frequency coefficient and spatial weight factor in the model are adjusted to make the collection sequence generated by the model closer to the actual environmental changes.

[0025] In the construction process of the data transmission model, its action mechanism is to clearly describe the data interaction process between the edge and the cloud. A transmission model that supports bidirectional communication is selected, which needs to be compatible with the communication protocols of edge computing nodes and cloud servers. Delay, packet loss, and bandwidth balance rules for data transmission are established. Different types of data are set with transmission priorities, such as key physiological parameters such as leaf color change and flower opening degree as high priority, environmental parameters such as temperature and humidity as medium priority, and sensor state information as low priority. When the network bandwidth is limited, high-priority data is transmitted first, and low-priority data is compressed to reduce transmission delay and packet loss, simulating the stability of data transmission. The delay time, packet loss rate, and bandwidth occupancy rate recorded during actual transmission are used to calibrate the data transmission model. The transmission performance of different priority data is analyzed, and the data packet fragmentation strategy and retransmission mechanism in the model are adjusted to ensure stable data transmission performance under different network conditions.

[0026] In the construction process of the threshold discrimination model, the mechanism is defined to accurately describe the correlation discrimination process between environmental parameters and health indicators. A discrimination model based on rule-based reasoning is selected, which needs to integrate the discrimination logic of multi-dimensional parameters. The parameter range and correlation rules corresponding to the threshold discrimination are established. For environmental characteristic parameters, specific threshold intervals are set, for example, the suitable interval of temperature is set to 18-24°C. When the temperature is lower than 18°C, the health indicators related to stem growth retardation are associated; when the temperature is higher than 24°C, the health indicators related to leaf wilting are associated; the suitable interval of humidity is set to 55%-75%, when the humidity is lower than 55%, the health indicators related to flower opening degree decrease are associated; when the humidity is higher than 75%, the health indicators related to leaf color darkening are associated; the suitable interval of light intensity is dynamically adjusted according to the growth stage, which is set to 10000-15000 lux at the germination stage and 15000-20000 lux at the flowering stage. When it exceeds the corresponding interval, different health indicators change are associated. The threshold discrimination model is calibrated through the actual monitoring of physiological characteristic parameters such as leaf color and stem diameter. The health status discrimination results output by the model are compared with the actual growth state of the Chinese flowering crabapple, and the weight coefficients in the correlation rules are optimized to make the model more accurate in identifying the health status of the Chinese flowering crabapple through environmental parameter changes.

[0027] In the construction process of the edge intelligent control module, the state space is first defined as the growth parameter vector collected in real time, which includes environmental parameters and physiological parameters of Chinese flowering crabapple. Environmental parameters include temperature, humidity and light intensity, among which temperature is in Celsius, humidity is in percentage and light intensity is in lux; physiological parameters include leaf color, stem diameter and flower opening degree, leaf color is represented by RGB value collected by color sensor, stem diameter is in millimeters, and flower opening degree is converted to a value between 0 and 1 through image recognition. These parameters together constitute a complete information set describing the growth state of Chinese flowering crabapple and the environment it is in, providing basic data for control decision.

[0028] The action space is defined as the adjustable parameter set of environmental control equipment, including light supplement duration adjustment, irrigation amount increment and continuous control amount of ventilation frequency. Light supplement duration adjustment is in minutes, positive value means increasing light supplement time, negative value means reducing; irrigation amount increment is in liters, positive value means increasing irrigation amount, negative value means reducing; ventilation frequency is in hours, the value represents the change amount of ventilation frequency per hour. The boundary of action space is determined according to the physical performance of the equipment, for example, the range of light supplement duration adjustment is limited to-60 to 60 minutes to avoid exceeding the safe range of equipment operation.

[0029] An Actor-Critic network architecture containing a communication reliability check gate is constructed. The Actor network consists of an input layer, a hidden layer and an output layer. The input layer receives the parameter vector of the state space, the hidden layer extracts features through a multi-layer perceptron, and the output layer generates continuous control actions. Each output node corresponds to a controllable parameter in the action space, and the output value is normalized and mapped to the actual control range. The Critic network also contains an input layer, a hidden layer and an output layer. The input layer receives the state space parameters and the action vector generated by the Actor network, the hidden layer processes the fused features, and the output layer designs a multi-objective reward function and evaluates the multi-objective reward.

[0030] The design of the multi-objective reward function combines the environmental suitability, health index and control energy consumption through dynamic weight coefficients to construct a suitability improvement term, a health degradation penalty term and an energy consumption gradient penalty term. The environmental suitability is calculated by the matching degree of environmental parameters and the optimal interval, the health index is determined based on the change trend of physiological parameters, and the control energy consumption is calculated according to the device operating power and duration. The health degradation penalty increases nonlinearly with the deviation of temperature and humidity from the optimal interval. Specifically, when the temperature and humidity are within the optimal interval, the penalty term is 0. When the deviation is small, the penalty strength increases slowly. When the deviation exceeds a certain threshold, the penalty strength increases rapidly.

[0031] The expression of the multi-objective reward function is:

[0032] where R represents the multi-objective reward value, α, β, γ are the dynamic weight coefficients of the environmental suitability, health degradation penalty term and energy consumption gradient penalty term respectively, F represents the suitability improvement term, P represents the health degradation penalty term, and G represents the energy consumption gradient penalty term.

[0033] The implementation of the communication reliability check gate is to add a constraint check module to the output layer of the Critic network. The module contains a small neural network, which takes the predicted Q value of the Critic network output and the communication reliability coefficient as input. The communication reliability coefficient is calculated based on the delay, packet loss rate and bandwidth utilization of data transmission, and its range is between 0 and 1. The higher the value, the more reliable the communication. The neural network layer performs nonlinear fusion on the input through the activation function, adjusts the weight of the predicted Q value, and reduces the weight of the predicted Q value when the communication reliability coefficient is low, and vice versa. Finally, the output is the action value evaluation result corrected by the communication constraint, so that the network evaluation can still be reasonable when the communication condition changes.

[0034] The training process of the network adopts a combination of offline pre-training and online fine-tuning. In the pre-training stage, simulated data generated by a virtual environment is used to enable the network to initially master basic control rules. In the online fine-tuning stage, the network parameters are continuously updated using actual collected data. Through the gradient descent algorithm, the evaluation error of the Critic network is minimized, and the Actor network is guided to generate more optimal control actions. During the training process, network parameters are saved regularly to avoid model performance degradation caused by abnormal data, ensuring that the edge intelligent control module can operate stably and output effective control instructions.

[0035] In the adjustment method of the dynamic weight coefficient, the monitoring data within the preset period is input into the stage recognition model. The model analyzes the change law of the environmental parameters and the characteristic value of the physiological parameters to determine the current growth stage of the Malus halliana. The monitoring data includes the temperature fluctuation curve, humidity change trend, light intensity distribution in the past 72 hours, and the RGB value change of leaf color, stem diameter growth data, and daily records of flower opening degree. The stage recognition model compares these data with the preset characteristic templates of each stage and outputs the current stage, including the germination stage, flowering stage, and fruiting stage.

[0036] In the germination stage, the root system and stems and leaves of Malus halliana begin to develop, and the sensitivity to environmental conditions is higher. At this time, the dynamic weight coefficient corresponding to the environmental suitability is increased to the preset weight threshold value, which is set according to the dependence of the germination stage on the environment. According to the constraint condition that the sum of all dynamic weight coefficients is 1, the dynamic weight coefficients corresponding to the health degradation penalty term and the energy consumption exceeding gradient penalty term are randomly allocated. For example, the weight coefficient of environmental suitability is set to 0.7, and the remaining 0.3 weight is randomly allocated between the health degradation penalty term and the energy consumption exceeding gradient penalty term. It may appear that the health degradation penalty term accounts for 0.1 and the energy consumption exceeding gradient penalty term accounts for 0.2, or each accounts for 0.15, to prioritize the environmental parameters in the suitable interval.

[0037] In the flowering stage, the flowers of the Chinese flowering crabapple gradually open, and the health status directly affects the quality of flowering. The dynamic weight coefficients corresponding to the health deterioration penalty term and the energy consumption exceeding gradient penalty term are dynamically allocated according to the health index quota. The health index quota is determined according to the ideal range of parameters such as flower opening degree and petal color. When the actual health index is higher than the quota, the weight of the health deterioration penalty term is reduced, and the weight of the energy consumption exceeding gradient penalty term is increased. When the actual health index is lower than the quota, the weight of the health deterioration penalty term is increased, and the weight of the energy consumption exceeding gradient penalty term is reduced. For example, when the flower opening degree reaches 80% or more, the weight of the health deterioration penalty term is reduced from 0.4 to 0.2, and the weight of the energy consumption exceeding gradient penalty term is increased from 0.2 to 0.4. When the flowers show signs of wilting and the opening degree drops below 50%, the weight of the health deterioration penalty term increases to 0.5, and the weight of the energy consumption exceeding gradient penalty term decreases to 0.1.

[0038] In the fruiting stage, the Chinese flowering crabapple begins to bear fruit, and the development of the fruit requires a stable environment and continuous nutrient supply. At this time, an emergency adjustment factor is introduced, which is determined according to the rate of change of parameters such as fruit growth rate and fruit color. When the fruit growth index is within the normal range, the emergency adjustment factor is 1, and it has no effect on the weight of the energy consumption exceeding gradient penalty term. When the fruit growth appears to be stagnant or the color is abnormal, the emergency adjustment factor is greater than 1, which increases the weight of the energy consumption exceeding gradient penalty term to limit unnecessary energy consumption. When extreme weather is encountered that may affect fruit development, the emergency adjustment factor is less than 1, which reduces the weight of the energy consumption exceeding gradient penalty term to allow appropriate increase in energy consumption to maintain environmental stability.

[0039] In the calibration method of the health deterioration penalty term, gradient experiments are conducted in different temperature and humidity intervals of the Chinese flowering crabapple. The temperature gradient is set to 10℃, 15℃, 20℃, 25℃, 30℃, and the humidity gradient is set to 40%, 50%, 60%, 70%, 80%, 90%. Each temperature and humidity combination forms an experimental interval. The growth status of the Chinese flowering crabapple is continuously monitored in each interval, and the number of wilted leaves is recorded daily. The wilted leaf rate is calculated, which is the proportion of wilted leaf number to total leaf number. By analyzing the wilted leaf rate at each interval point, the temperature and humidity influence factor is calibrated, which increases with the increase of the degree of deviation of temperature and humidity from the optimal interval. The mapping relationship between the temperature and humidity deviation degree and the health deterioration penalty gain coefficient is established. The temperature and humidity deviation degree is calculated by the difference between the actual temperature and humidity and the center value of the optimal interval, and the health deterioration penalty gain coefficient is determined according to the change trend of the wilted leaf rate. When the temperature and humidity deviation degree is 0, the gain coefficient is 0. When the deviation degree increases by 1 unit, the gain coefficient increases accordingly according to the mapping relationship, so as to calibrate the health deterioration penalty term.

[0040] Example 5: The optimization objectives include maximizing the suitability of environmental parameters, maximizing the health index of crabapple flowers, and minimizing the energy consumption of regulation. Maximizing the suitability of environmental parameters requires that parameters such as temperature, humidity, and light intensity be as close as possible to the optimal range for crabapple growth. This range is determined according to the characteristics of crabapple flowers at different growth stages. For example, during the flowering stage, the optimal temperature range may be set at 20-25°C, humidity at 65%-75%, and light intensity at 15,000-20,000 lux. Maximizing the health index of crabapple flowers focuses on the status of physiological parameters such as leaf color, stem diameter, and flower openness. The leaf color needs to remain bright and uniform, the stem diameter needs to show a steady growth trend, and the flower openness needs to be within the normal range of this growth stage. Minimizing the energy consumption of regulation requires that in the process of achieving the first two goals, the operating time and power of lighting equipment, irrigation equipment, ventilation equipment, etc. should be reasonably controlled to reduce unnecessary energy consumption.

[0041] To implement this hierarchical and progressive environmental control approach, pre-training within the constraints of the collaborative model is first performed in a virtual environment. This virtual environment uses a computer program to simulate various possible crabapple flower growth scenarios, including seasonal climate conditions, sudden weather changes, and differences in soil fertility. The collaborative model comprises an edge IoT perception model and a crabapple flower health assessment model. During pre-training, the edge IoT perception model simulates the collection and processing of environmental and physiological data, while the crabapple flower health assessment model assesses the flower's health based on this data. Under the constraints of both models, a large number of simulated control experiments are conducted to generate an initial control strategy library. This library contains preliminary control solutions for different scenarios, such as cooling solutions for hot weather and supplemental lighting solutions for insufficient light.

[0042] Data-driven policy fine-tuning is achieved through a dual experience replay pool. This dual experience replay pool consists of a data compliance pool and an actual optimization pool. The data compliance pool stores experience samples that meet preset constraints, such as complete data acquisition timestamps, no transmission loss, and parameter values ​​within a reasonable range. The actual optimization pool stores valid samples during operation. Valid samples are those that improve environmental parameter suitability, increase health indexes, or reduce energy consumption during actual control. The sampling priority of the dual experience replay pool is dynamically determined by a weighted fusion of the data compliance score and the reward value. The data compliance score is calculated based on the degree to which a sample meets the constraints. Samples that fully meet the constraints receive the highest score, followed by samples with minor deviations. Non-compliant samples are excluded. The reward value is calculated based on a multi-objective reward function and reflects the contribution of the corresponding control action to the optimization objective. High-priority samples are selected first for policy fine-tuning. Through continuous iteration, the control strategy gradually adapts to the characteristics of the actual growth environment.

[0043] The action space constraint is dynamically relaxed according to the operation stability index. The operation stability index is embodied by the fluctuation rate of the environmental parameter, and the fluctuation rate of the environmental parameter is calculated as the ratio of the parameter change value in the last two control periods to the optimal interval of the parameter. When the fluctuation rate of the environmental parameter in the last N control periods is not higher than the preset fluctuation rate threshold, it indicates that the current environment is in a relatively stable state, and the constraint parameter is updated at this time. The update method is to adaptively relax the action space constraint parameter according to the operation stability index. The action space constraint parameter includes the light supplement time length adjustment amount, the irrigation amount increment, and the upper and lower limits of the ventilation frequency. When the performance index reaches the preset index threshold, the allowed control amount change range is expanded in proportion. For example, the initial range of the light supplement time length adjustment amount is -30 to 30 minutes, and when the operation stability index meets the standard for 5 consecutive periods, the range is expanded to -45 to 45 minutes; the initial range of the irrigation amount increment is -5 to 5 liters, and after meeting the standard, it is expanded to -7.5 to 7.5 liters; the initial range of the ventilation frequency is -2 to 2 times / hour, and after meeting the standard, it is expanded to -3 to 3 times / hour. In this way, the control is more flexible to adapt to possible subtle environmental changes, while avoiding large-scale control when the environment is unstable, and ensuring the stability of the growth environment of the Chinese hibiscus.

[0044] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0045] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for controlling and monitoring the environment of crabapple flowers based on edge IoT, characterized in that: include: S1: Build a dual-driven model that is coordinated by the edge IoT perception model and the crabapple flower health assessment model; S2: Use a dual-drive model to monitor the growth process of crabapple flowers and design an edge-end intelligent control module based on the dual-drive model; S3: Optimize the control strategy according to the optimization target, perform hierarchical progressive environmental control, and adjust the environmental control equipment.

2. The method for controlling and monitoring the environment of crabapple flowers based on edge IoT according to claim 1, characterized in that: The steps of constructing the edge IoT perception model include: Step S101: Analyze the growth environment characteristics of the crabapple flower and determine characteristic parameters; Characteristic parameters include environmental characteristic parameters and physiological characteristic parameters; environmental characteristic parameters include temperature, humidity and light intensity, and physiological characteristic parameters include leaf color, stem diameter and flower openness; Step S102: Analyze the mechanism of crabapple flower environmental regulation and health monitoring, and build an edge IoT perception model; The edge IoT perception model consists of an environment collection model, a data transmission model, and a threshold discrimination model; Step S103: using numerical calculation methods to simulate the growth process of crabapple flowers, and optimizing the edge IoT perception model based on the simulation results; The equations of each model in the edge IoT perception model are converted into discrete mathematical forms; initial conditions are set; numerical calculation methods are used to solve the discretized equations to simulate the evolution of environmental data at each stage of the crabapple flower growth process; the simulation results are post-processed; and the model parameters of the edge IoT perception model are optimized based on the difference between the simulation results after post-processing and the characteristic parameters.

3. The method for controlling and monitoring the environment of crabapple flowers based on edge IoT according to claim 2, characterized in that: The construction process of the environmental acquisition model is as follows: determining the working mechanism of the environmental acquisition model, which is to describe the real-time acquisition process of environmental parameters; selecting the environmental acquisition model; establishing the time, space and frequency balance rules for data acquisition to simulate the continuous acquisition process of environmental parameters; and calibrating the environmental acquisition model using the corresponding characteristic parameters. The data transmission model construction process is as follows: determining the mechanism of the data transmission model, which describes the data interaction process between the edge and the cloud; selecting the data transmission model; establishing data transmission delay, packet loss and bandwidth balance rules to simulate the stability of data transmission; and calibrating the data transmission model using corresponding characteristic parameters. The construction process of the threshold discrimination model is as follows: determining the action mechanism of the threshold discrimination model, wherein the action mechanism of the threshold discrimination model is to describe the association discrimination process between environmental parameters and health indicators; Select a threshold discrimination model; establish a parameter range and association rules corresponding to the threshold discrimination; and calibrate the threshold discrimination model through corresponding characteristic parameters.

4. The method for controlling and monitoring the environment of crabapple flowers based on edge IoT according to claim 3 is characterized in that: The method for constructing the edge intelligent control module includes: The state space is defined as the growth parameter vector collected in real time, including environmental parameters and physiological parameters of the crabapple flower; The action space is defined as the set of adjustable parameters of the environmental control equipment, including the continuous control of the light duration adjustment, irrigation increment and ventilation frequency; Build an Actor-Critic network architecture with a communication reliability check gate, where the Actor network is used to generate continuous control actions, and the Critic network is used to design multi-objective reward functions and evaluate multi-objective rewards; The design method of the multi-objective reward function includes: By combining the dynamic weight coefficient with the environmental suitability, health index and regulated energy consumption, a suitability improvement item, a health degradation penalty item and a gradient penalty item for exceeding the energy consumption standard are constructed. A multi-objective reward function is obtained based on the suitability improvement item, the health degradation penalty item and the gradient penalty item for exceeding the energy consumption standard. Among them, the intensity of the health degradation penalty increases nonlinearly as the temperature and humidity deviate from the optimal range.

5. The method for controlling and monitoring the environment of crabapple flowers based on edge IoT according to claim 4 is characterized in that: The implementation method of the communication reliability check gate includes: adding a constraint check module to the output layer of the critic network, wherein the constraint check module performs nonlinear fusion of the predicted Q value and the communication reliability coefficient through the neural network layer, and outputs the action value evaluation result corrected by the communication constraint.

6. The method for controlling and monitoring the environment of crabapple flowers based on edge IoT according to claim 4, characterized in that: The method for adjusting the dynamic weight coefficient includes: The monitoring data within a preset time period is used as the input of the stage recognition model to obtain the current stage, which includes the bud stage, flowering stage and fruit stage; in the bud stage, the dynamic weight coefficient corresponding to the environmental suitability is increased to the preset weight threshold, and the dynamic weight coefficients corresponding to the health degradation penalty item and the energy consumption exceeding gradient penalty item are randomly allocated according to the constraint that the sum of all dynamic weight coefficients is 1; in the flowering stage, the dynamic weight coefficients corresponding to the health degradation penalty item and the energy consumption exceeding gradient penalty item are dynamically allocated according to the health indicator quota; in the fruit stage, an emergency adjustment factor is introduced to adjust the dynamic weight coefficient corresponding to the energy consumption exceeding gradient penalty item.

7. The method for controlling and monitoring the environment of crabapple flowers based on edge IoT according to claim 4, characterized in that: The calibration method of the health degradation penalty item includes: A gradient experiment was conducted on crabapple flowers in different temperature and humidity ranges. The leaf wilting rate at each interval point was collected. The temperature and humidity influencing factors were calibrated based on the leaf wilting rate at each interval point. A mapping relationship between the temperature and humidity deviation and the health degradation penalty gain coefficient was established. The health degradation penalty term was obtained by calibrating the mapping relationship between the temperature and humidity deviation and the health degradation penalty gain coefficient.

8. The method for controlling and monitoring the environment of crabapple flowers based on edge IoT according to claim 7, characterized in that: The optimization objectives of the optimization and control strategy include maximizing the suitability of environmental parameters, maximizing the health index of crabapple flowers, and minimizing the control energy consumption.

9. The method for controlling and monitoring the environment of crabapple flowers based on edge IoT according to claim 8, characterized in that: The method for performing hierarchical progressive environmental control includes: Conduct pre-training under collaborative model constraints in a virtual environment to generate an initial control strategy library; Data-driven policy fine-tuning through dual experience replay pools; Dynamically relax action space constraints based on operational stability indicators; The dual experience replay pool includes a data compliance pool and an actual optimization pool; the data compliance pool is used to store experience samples that meet preset constraints; the actual optimization pool is used to store samples that are valid during operation; wherein, the sampling priority of the dual experience replay pool is dynamically determined by the weighted fusion result of the data compliance score and the reward value.

10. The method for controlling the environment and monitoring the health of crabapple flowers based on edge IoT according to claim 9, characterized in that: The method for dynamically relaxing action space constraints includes: When the environmental parameter volatility for N consecutive control cycles is not higher than the preset volatility threshold, the constraint parameter is updated. The method for updating the constraint parameter includes: The action space constraint parameters are adaptively relaxed and adjusted according to the operation stability index. When the performance index reaches the preset index threshold, the range of the allowable control amount is proportionally expanded.

Images