Pepper breeding isolation shed control system based on internet of things

Abstract

The present application relates to the technical field of pepper breeding regulation, and discloses a pepper breeding isolation shed control system based on the Internet of Things. The system comprises an environmental data acquisition module, a user behavior analysis module, an isolation shed partition module, a dynamic baseline generation module, an abnormal interaction detection module, a root node decision module and a recommendation chain construction module. The environmental data acquisition module acquires temperature, humidity, light intensity and carbon dioxide concentration data in the shed in real time; the user behavior analysis module analyzes user historical regulation instructions; the isolation shed partition module divides the regulation sub-regions according to environmental differences; the dynamic baseline generation module generates dynamic environmental parameter baselines for each sub-region; the abnormal interaction detection module identifies abnormal environmental fluctuations and instructions; the root node decision module filters the most significant root node area of fluctuations; and the recommendation chain construction module constructs a multi-level regulation recommendation chain. The system improves the accuracy and intelligence of the isolation shed environmental regulation and meets the pepper breeding requirements.

Description

Technical Field

[0001] This invention relates to the field of chili breeding regulation technology, specifically to a chili breeding isolation shed control system based on the Internet of Things. Background Technology

[0002] In chili pepper breeding, isolation greenhouses serve as crucial facilities for ensuring breeding purity and stability. The precise control of their internal environmental parameters directly impacts the breeding process and the quality of the results. Traditional chili pepper breeding isolation greenhouse environmental management relies heavily on manual operation. Staff must periodically enter the greenhouse to collect data such as temperature, humidity, light intensity, and carbon dioxide concentration using handheld devices. This process is not only time-consuming and labor-intensive but also prone to data delays and inaccuracies due to human error. Especially in large-scale breeding scenarios, a single environmental monitoring and control model is insufficient to meet the diverse growth environment requirements of different chili pepper varieties in various regions. In some areas, unsuitable environmental parameters may lead to slow growth, decreased disease resistance, and other problems, thus affecting overall breeding efficiency.

[0003] With the development of agricultural IoT technology, some isolation greenhouses have begun to introduce automated monitoring equipment, but existing systems still have significant limitations. Most systems can only achieve fixed threshold alarms for single environmental parameters, lacking the ability to deeply analyze historical data and dynamically adjust them. They cannot generate adaptive environmental parameter standards based on the needs of different growth stages of chili peppers. Furthermore, existing systems typically treat the isolation greenhouse as a whole for regulation, ignoring environmental differences caused by factors such as equipment layout, ventilation conditions, and light angles in different locations within the greenhouse, leading to uneven regulation effects. In addition, at the user operation level, the system cannot effectively analyze historical control commands, making it difficult to identify unreasonable commands or provide targeted control suggestions. This makes it difficult for users to make quick and scientific decisions when facing complex environmental fluctuations, further restricting the intelligence and precision of environmental regulation in isolation greenhouses and failing to fully meet the demands of modern chili pepper breeding for efficient and precise environmental management. Summary of the Invention

[0004] The purpose of this invention is to provide an Internet of Things-based control system for chili breeding isolation sheds to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an Internet of Things (IoT)-based control system for chili breeding isolation sheds, the system comprising:

[0006] The environmental data acquisition module is used to acquire real-time data on temperature, humidity, light intensity, and carbon dioxide concentration inside the isolation shed.

[0007] The user behavior analysis module is used to receive and analyze historical data of control commands input by users through terminal devices, and extract command frequency, command type and command time interval;

[0008] The isolation shed zoning module divides the isolation shed into multiple control sub-regions with different environmental parameters based on the differences in data obtained by the environmental data acquisition module.

[0009] The dynamic baseline generation module generates dynamic environmental parameter baselines for each control sub-region based on historical data from the environmental data acquisition module and instruction data from the user behavior analysis module.

[0010] The abnormal interaction detection module compares the deviation between real-time environmental data and the baseline of dynamic environmental parameters to identify abnormal environmental fluctuations and abnormal user commands.

[0011] The root node decision module selects the control sub-region with the most significant environmental fluctuations as the root node region based on the output of the abnormal interaction detection module.

[0012] The recommendation chain construction module, centered on the root node region, combines the real-time environmental data similarity of other regulation sub-regions to construct a multi-level environmental regulation recommendation chain.

[0013] Preferably, the user behavior analysis module specifically includes:

[0014] The historical data of user commands is divided into stable command periods and variable command periods according to the time series.

[0015] Extract the average execution frequency of each instruction type during stable instruction periods, and extract the time and direction of instruction type mutations during variable instruction periods;

[0016] The set of instruction types to be analyzed is selected based on the difference coefficient of instruction frequency between stable and variable periods.

[0017] Preferably, the dynamic baseline generation module specifically includes:

[0018] For each control sub-region, the ratio of the standard deviation of environmental parameters during the stable command period to the standard deviation of environmental parameters during the changing command period is calculated and used as the baseline dynamic adjustment factor.

[0019] The average environmental parameters for a consecutive preset period before the moment of sudden change in user command type are used as the temporary environmental baseline for this control sub-region.

[0020] The temporary environmental baseline is weighted and corrected based on the baseline dynamic adjustment factor to generate the final dynamic environmental parameter baseline.

[0021] Preferably, the abnormal interaction detection module specifically includes:

[0022] For each control sub-region, the Euclidean distance between the current environmental parameters and the dynamic environmental parameter baseline is calculated in real time;

[0023] When the Euclidean distance exceeds a preset threshold, extract the user instruction type and execution timestamp corresponding to that moment.

[0024] If the interval between the user command execution timestamp and the start time of environmental parameter fluctuations exceeds a preset delay threshold, it is marked as an abnormal interaction event.

[0025] Preferably, the root node decision module specifically includes:

[0026] Calculate the maximum Euclidean distance among abnormal interaction events in all control sub-regions;

[0027] The control sub-region corresponding to the maximum value is taken as the root node region, and the remaining regions are arranged in descending order of Euclidean distance as candidate nodes.

[0028] Record the environmental parameter similarity matrix between the root node region and each candidate node region at the time of abnormal fluctuation.

[0029] Preferably, the recommendation chain construction module specifically includes:

[0030] Starting from the root node region, the candidate node with the highest environmental parameter similarity is selected as the first-level recommended node;

[0031] Based on the primary recommended nodes, candidate nodes with the second highest similarity are selected as secondary recommended nodes;

[0032] The recommendation chain is constructed by iteratively building multiple layers until all candidate nodes are covered or the similarity is lower than the preset association threshold.

[0033] Preferably, the system further includes: a covariance correction module, which dynamically corrects the noise covariance matrix of environmental control commands based on the real-time data fluctuation amplitude of the environmental data acquisition module and the response delay of user commands, specifically including:

[0034] Calculate the first-order difference absolute value sequence of the real-time sampling sequence of environmental parameters, and take the maximum value in the sequence as the fluctuation intensity benchmark;

[0035] The delay time from the generation to the execution of user instructions is statistically analyzed, and the product of the delay time and the fluctuation intensity benchmark is calculated as the correction weight.

[0036] The initial values ​​of the noise covariance matrix are dynamically updated using modified weights.

[0037] Preferably, the system further includes: a cross-layer association module, which integrates the physical layer data of the environmental data acquisition module and the instruction layer data of the user behavior analysis module to identify cross-layer abnormal association patterns, specifically including:

[0038] Extract time period markers for abnormal fluctuations in physical layer environmental parameters;

[0039] Retrieve all abnormal instruction records in the user instruction layer within the same time period;

[0040] The Pearson correlation coefficient between the fluctuation range of environmental parameters and the number of abnormal commands was calculated as an indicator of the cross-layer correlation strength.

[0041] Preferably, the system further includes: a source tracing execution module, which, based on the output of the cross-layer correlation module, traces the root cause of abnormal environmental fluctuations and generates a targeted control instruction sequence, specifically including:

[0042] For abnormal events where the cross-layer correlation strength exceeds a preset threshold, the trajectory of environmental parameter changes in the root node region before the abnormal fluctuation is traced back; the moment when the deviation from the dynamic environmental parameter baseline first appears in the matching trajectory is taken as the moment when the root cause occurs; based on the gradient change direction of the environmental data collected before and after the moment when the root cause occurs, a reverse control instruction sequence is generated.

[0043] Preferably, the logic for generating the reverse control instruction sequence includes:

[0044] If the gradient of environmental parameters changes in a continuously increasing direction, then a control command that gradually decreases will be generated.

[0045] If the gradient of environmental parameters changes in a continuously decreasing direction, then progressively increasing control commands are generated.

[0046] The magnitude of each instruction level is linearly proportional to the deviation of environmental parameters at the time the root cause occurs.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] This IoT-based control system for chili breeding isolation sheds effectively solves many problems in traditional isolation shed environmental control through multi-module collaborative operation, providing a more efficient and precise environmental management solution for chili breeding. The environmental data acquisition module can acquire real-time data on temperature, humidity, light intensity, and carbon dioxide concentration within the isolation shed, eliminating reliance on manual data collection, avoiding human error, and ensuring the timeliness and accuracy of environmental data. This allows staff to monitor the dynamic environment within the shed in real time, providing a reliable data foundation for subsequent control decisions.

[0049] The user behavior analysis module analyzes users' historical control commands, extracting key information such as command frequency, type, and time interval. This clearly reveals users' operating habits and control preferences, while also identifying potential unreasonable operating patterns. This helps users standardize their operations, reduce environmental fluctuations caused by misoperation, and improve the scientific nature of overall control. The isolation greenhouse zoning module divides the isolation greenhouse into multiple control sub-regions based on differences in environmental data. This breaks the limitations of traditional overall control, fully considering the environmental heterogeneity of different locations within the greenhouse. This allows the control of each sub-region to better suit the growth needs of the chili varieties in that region, avoiding the problem of environmental unsuitability in some areas due to uniform control, and improving the targeting and balance of environmental control.

[0050] The dynamic baseline generation module generates dynamic environmental parameter baselines for each control sub-region based on historical environmental data and user command data, rather than using fixed threshold standards. This dynamic baseline can adaptively adjust according to the needs of different growth stages of chili peppers and long-term environmental trends, making the environmental parameter standards more consistent with actual growth patterns. When environmental parameters deviate from the baseline, it can issue timely warnings, providing a clear direction for precise control. The anomaly detection module compares the deviation between real-time environmental data and the dynamic baseline, quickly identifying abnormal environmental fluctuations and detecting abnormal user commands. This dual monitoring mechanism effectively reduces the impact of environmental anomalies on chili pepper growth and also reduces interference from unreasonable commands on system control, improving system stability.

[0051] Upon detecting anomalies, the root node decision module can select the most significantly fluctuating environmental sub-region as the root node region. This helps staff quickly pinpoint the core problem area, avoiding blind investigation in complex multi-region environments, saving time in problem localization, and improving emergency response efficiency. The recommendation chain construction module, centered on the root node region, combines real-time environmental data similarity from other sub-regions to construct a multi-level control recommendation chain. This provides users with clear and orderly control path suggestions. Users do not need to analyze complex environmental relationships themselves; they can gradually complete precise control of each region based on the recommendation chain, reducing operational difficulty while ensuring the consistency and effectiveness of control measures. This further enhances the intelligence level of environmental control in the isolation shed, creating a more suitable growth environment for chili pepper breeding and facilitating the efficient advancement of breeding work. Attached Figure Description

[0052] Figure 1 This is a timing diagram of the Internet of Things-based chili breeding isolation shed control system described in this invention;

[0053] Figure 2 A schematic diagram illustrating the working principle of the user behavior analysis module;

[0054] Figure 3This is a schematic diagram illustrating the working principle of the abnormal interaction detection module. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Please see Figure 1 This invention provides an Internet of Things-based control system for chili breeding isolation sheds. The system includes: an environmental data acquisition module, a user behavior analysis module, an isolation shed zoning module, a dynamic baseline generation module, an abnormal interaction detection module, a root node decision module, and a recommendation chain construction module.

[0057] An environmental data acquisition module, deployed within a sensor network inside the isolation shed, acquires real-time data on temperature, humidity, light intensity, and carbon dioxide concentration. This data is transmitted to the central processing unit via an IoT protocol. A user behavior analysis module receives historical data of control commands input by users through terminal devices, analyzing command frequency, type, and time intervals to extract user operation patterns. An isolation shed zoning module divides the isolation shed into multiple control sub-regions based on differences in the data acquired by the environmental data acquisition module. Each sub-region has different environmental parameter characteristics, such as areas with higher temperatures or lower humidity. A dynamic baseline generation module generates a dynamic environmental parameter baseline for each control sub-region based on historical environmental data and user command data. This baseline adaptively adjusts over time and with user behavior. An abnormal interaction detection module compares the deviation between real-time environmental data and the dynamic environmental parameter baseline to identify abnormal environmental fluctuations and abnormal user commands. A root node decision module, based on the output of the abnormal interaction detection module, selects the control sub-region with the most significant environmental fluctuations as the root node region. A recommendation chain construction module, centered on the root node region and combining the similarity of real-time environmental data from other control sub-regions, constructs a multi-level environmental control recommendation chain to optimize the overall control strategy.

[0058] Example 1: See Figure 2The collaborative operation of the user behavior analysis module and the dynamic baseline generation module forms the core of the system's understanding of user operation patterns and the establishment of adaptive environmental benchmarks. The user behavior analysis module first processes historical data of control commands from terminal devices. This data is stored in time-series format, including command type, execution timestamp, and parameter adjustment values. The module uses a sliding window algorithm to segment the time series, with the window size dynamically adjusted according to the data sampling frequency, thereby identifying stable command periods and variable command periods. Stable periods are characterized by a uniform distribution of command types and relatively fixed execution intervals, while variable periods exhibit abrupt changes in command frequency or type switching. Within stable command periods, the module calculates the average execution frequency of each command type, such as the average number of triggers for temperature control commands over 24 consecutive hours, and records the temporal distribution patterns of these commands. Within variable command periods, the module uses abrupt change detection algorithms (such as variance-based outlier analysis) to locate the moment of command type abrupt change and records the direction of the change (e.g., a sudden switch from an increase in temperature command to a decrease in humidity command).

[0059] The module further calculates the frequency difference coefficient between stable and variable periods. This coefficient is derived by comparing the frequency variance ratio of the two periods, thereby filtering out the set of instruction types to be analyzed. These sets typically include high-frequency instructions (such as frequent adjustments to illumination) and low-frequency but significantly abrupt instructions (such as occasional adjustments to carbon dioxide concentration). The dynamic baseline generation module generates dynamic environmental parameter baselines for each control sub-region based on the output of the user behavior analysis module. The module first calculates the standard deviation of environmental parameters during stable instruction periods, reflecting the fluctuation range of parameters during periods of stable user operation. Simultaneously, it calculates the standard deviation of environmental parameters during variable instruction periods, capturing the degree of variability of parameters during periods of user intervention. The ratio of these two values ​​serves as a baseline dynamic adjustment factor, quantifying the sensitivity of environmental parameters to changes in user behavior. The module then extracts the mean environmental parameters for a consecutive preset period before the abrupt change in user instruction type. The preset period length can be adjusted according to system configuration (e.g., 30 minutes before the abrupt change). This mean serves as the temporary environmental baseline for that control sub-region, representing a relatively stable environmental state before the abrupt change. The block uses a baseline dynamic adjustment factor to weight and correct the temporary environmental baseline. The weighting process involves multiplying the temporary baseline by the adjustment factor and adding the exponential moving average of historical environmental data to generate the final dynamic environmental parameter baseline. This baseline not only reflects historical environmental trends but also incorporates the impact of sudden changes in user behavior, enabling the baseline to adaptively track environmental changes.

[0060] The module iteratively updates the baseline value, ensuring it is always based on the latest user commands and environmental data, thus maintaining the accuracy and real-time performance of the baseline. The interaction between the user behavior analysis module and the dynamic baseline generation module is achieved through a data pipeline, which transmits command analysis results to the baseline generation stage in real time. For example, when the user behavior analysis module detects a sudden change in lighting commands, the dynamic baseline generation module immediately triggers the baseline recalculation process, quickly adjusting the baseline value using environmental data prior to the change. This design allows the system to tightly couple user operational intentions with changes in environmental state, providing a reliable benchmark reference for subsequent anomaly detection and decision-making. The module's implementation relies on a time-series database and a real-time computing framework, ensuring processing efficiency and low latency under large data volumes.

[0061] Example 2: See Figure 3 The anomaly interaction detection module and the root node decision module work together to identify environmental anomalies and determine priority areas for handling. The anomaly interaction detection module continuously receives real-time data streams from the environmental data acquisition module and baseline data from the dynamic baseline generation module. This module independently calculates the multidimensional differences between the current environmental parameters and the dynamic environmental parameter baseline for each control sub-region. The calculation process uses the Euclidean distance algorithm, which comprehensively evaluates the deviation of four dimensions: temperature, humidity, light intensity, and carbon dioxide concentration, forming a comprehensive anomaly index. When the Euclidean distance value of a sub-region exceeds a preset threshold, the system immediately triggers the anomaly event recording process. This threshold is set based on historical data statistics and typically uses a dynamic adjustment mechanism to adapt to changes in different seasons or crop growth stages. Simultaneously, the module automatically extracts the user command type and precise execution timestamp corresponding to the anomaly moment; this command data comes from the historical record database of the user behavior analysis module. The module further compares the time interval between the user command execution timestamp and the start time of the environmental parameter fluctuation. If this interval exceeds the system's preset delay threshold, the event is marked as an abnormal interaction event. The delay threshold is dynamically configured based on the device's response characteristics and network transmission delay. For example, it may be set to the second level in highly automated systems, while it may be extended to several minutes in scenarios where manual intervention is the main method.

[0062] The root node decision module receives data packets from all abnormal interaction events. This module first calculates the Euclidean distance values ​​of abnormal events across all control sub-regions and uses a sorting algorithm to find the maximum value. The control sub-region corresponding to the maximum value is determined as the root node region, representing the most urgent environmental anomaly in the current system. The remaining anomaly-affected sub-regions are then sorted in descending order of their Euclidean distance values ​​to form a candidate node list. Each node in the list includes its anomaly severity score and an occurrence timestamp. The module then constructs an environmental parameter similarity matrix between the root node region and each candidate node region at the time of the abnormal fluctuation. The similarity calculation uses a multi-dimensional feature matching algorithm, which considers both the absolute value similarity of the parameters and the consistency of their changing trends. The matrix storage employs sparse matrix compression technology to improve storage efficiency, where each element represents the environmental feature matching degree between the two regions at the time of the anomaly. When a new abnormal event is detected, the system immediately updates the root node decision and the similarity matrix. Asynchronous communication between modules is achieved through message queues, ensuring real-time responsiveness even in high-concurrency scenarios. The data persistence layer uses a time-series database to store all abnormal event records, supporting fast querying and historical backtracking analysis, providing the system with a complete abnormal audit trail.

[0063] At 10:22 AM on August 15, 2023, the environmental monitoring system in the chili breeding isolation shed in Zone 3 began recording abnormal data fluctuations. The environmental data acquisition module showed that the temperature parameter in the B-5 control sub-zone rose from 25.6℃ to 28.9℃ within 3 minutes, while the humidity dropped from 65% to 58%, and the carbon dioxide concentration also showed an abnormal increase. The abnormal interaction detection module immediately initiated the processing flow, calculating the Euclidean distance between the current environmental parameters and the baseline of the dynamic environmental parameters. This distance value reached 4.3, significantly exceeding the system's preset threshold of 2.5. The system automatically retrieved user operation records for that period and found a ventilation control command from a technician at 10:20:15, requesting an increase in ventilation volume in Zone B-5 by 30%. However, the environmental parameter fluctuation started at 10:22:08, with an interval of 113 seconds between the command execution time and the time, exceeding the system's set 90-second delay threshold. The module marked this event as an abnormal interaction event, recorded the command type as ventilation adjustment, and saved a complete environmental data snapshot.

[0064] Almost simultaneously, the system detected anomalies in other sub-regions: region C-2 experienced an abnormal temperature drop with a Euclidean distance of 3.1; region D-3 experienced an abnormal humidity increase with a distance of 3.8; and region E-1 experienced an abnormal carbon dioxide concentration with a distance of 2.9. All the timestamps of these anomalies occurred between 10:22 and 10:25, forming an anomaly cluster. The root node decision module began processing these anomalies, first calculating the Euclidean distance values ​​for each region, determining that region B-5 had the highest value of 4.3, and therefore designating it as the root node region. The remaining regions were then sorted in descending order of distance value: region D-3 (3.8), region C-2 (3.1), and region E-1 (2.9), forming a candidate node list. The module then calculates the similarity of environmental parameters between the root node region and each candidate node at the time of abnormal fluctuations. It finds that regions D-3 and B-5 have a similarity of 0.76 in temperature change trends, region C-2 has a similarity of 0.68 in humidity change patterns, and region E-1 shows a correlation of 0.52 in carbon dioxide concentration change.

[0065] The system packages all this data into an anomaly report, including the basis for root node determination, candidate node ranking results, and similarity matrix data. The report is sent to the recommendation chain building module via a message queue, while anomaly details are stored in a time-series database. Throughout the process, the system maintains real-time monitoring, updating environmental parameter readings every 30 seconds and dynamically adjusting Euclidean distance calculations and similarity assessments. The operator terminal receives an anomaly alert, indicating that area B-5 is marked as the highest priority processing object. Technicians can view the complete anomaly analysis report, including environmental parameter change curves, user command execution records, the root node determination process, and related area information. The system maintains the integrity of all raw data, supporting subsequent traceability analysis and providing sufficient evidence for regulatory decisions. From initial anomaly detection to final determination of processing priority, the entire process is completed within 2 minutes and 45 seconds, demonstrating the system's real-time response capability. The collaborative cooperation between modules ensures the timeliness and accuracy of anomaly event handling.

[0066] Example 3: The collaborative operation of the recommendation chain construction module and the covariance correction module. These two modules are jointly responsible for generating optimized control strategies and adjusting system response characteristics. The recommendation chain construction module takes the root node region and candidate node list output by the root node decision module as input. This module first calculates the similarity of environmental parameters between the root node region and each candidate node region at the current time. The similarity evaluation adopts the angle cosine metric method in multidimensional vector space, comprehensively considering the synchronous change characteristics of temperature, humidity, light, and carbon dioxide concentration. The module determines the candidate node with the highest similarity value as the first-level recommendation node, establishing a priority control association between the root node and the first-level node. This association means that the two regions may have the same root cause of the problem or characteristics that can be synergistically controlled. After the first-level recommendation node is determined, the module continues to select the candidate node with the second highest similarity to the first-level node as the second-level recommendation node. This process is achieved by iteratively comparing the average similarity between the remaining nodes and the selected nodes, ensuring the continuity of the recommendation chain expansion. The module recursively constructs a multi-level recommendation chain structure according to the principle of descending similarity. The depth of the chain depends on the preset association threshold. When the similarity of the newly added node is lower than the threshold, the expansion stops, and finally a tree-like recommendation topology is formed.

[0067] During the recommendation chain construction process, the module employs a dynamic weight adjustment mechanism. The similarity weight of each node is updated based on real-time data, and the weight calculation involves the instantaneous change rate of environmental parameters and the historical association strength. The module maintains a real-time updated priority queue to manage the node order in the recommendation chain, ensuring the stability of the chain structure even in high-frequency data update scenarios. All inter-node association data is stored in a relational database, supporting fast querying and dynamic reconstruction, providing the system with flexible strategy adjustment capabilities. The covariance correction module works in parallel. Based on the real-time data fluctuation characteristics of the environmental data acquisition module and the response latency characteristics of user commands, this module dynamically adjusts the parameters of the noise covariance matrix. The module first calculates the first-order difference absolute value sequence of the real-time sampling sequence of environmental parameters. This sequence reflects the intensity of change of each parameter in a short period of time. The maximum value is extracted from this sequence as the fluctuation intensity benchmark value, which characterizes the maximum disturbance level of the current system. The module also statistically analyzes the latency data from the generation to the actual execution of user commands. This data comes from the timestamp records of the user behavior analysis module, and the latency calculation is accurate to the millisecond level. The module multiplies the delay duration by the fluctuation intensity benchmark value to obtain the correction weight, which is used to quantify the impact of system response delay on the environmental control effect.

[0068] The noise covariance matrix is ​​updated using the following relationship:

[0069]

[0070] in: This represents the updated noise covariance matrix. Represents the initial noise covariance matrix. The system sensitivity coefficient, The corrected weights are calculated. This update process ensures that the matrix reflects the actual noise characteristics of the current system, improving the accuracy of state estimation.

[0071] The two modules exchange data via a shared memory area. The recommendation chain construction module updates the uncertainty factor in its similarity calculation using the matrix output by the covariance correction module, forming a closed-loop optimization system. The entire implementation adopts a microservice architecture, with each module deployed independently but maintaining data synchronization through a lightweight communication protocol, ensuring high availability and scalability of the system. Data stream processing uses an event sourcing model, with all state changes recorded in logs, supporting fault recovery and system backtracking analysis.

[0072] At 2:30 PM on September 20, 2023, the system detected multiple environmental anomalies within the isolation shed: In area B-2, the temperature abnormally rose to 31.5℃, and the humidity dropped to 55%; in area D-4, the carbon dioxide concentration abnormally surged to 1200 ppm; and in area F-1, the light intensity suddenly dropped to 300 lux. The root node decision module identified area B-2 as the primary processing target due to its Euclidean distance value of 4.8, significantly higher than other areas. The recommendation chain construction module immediately initiated the processing flow, first calculating the real-time environmental parameter similarity between area B-2 and the other abnormal areas. The system employed a cosine similarity algorithm in a multi-dimensional feature space to comprehensively analyze the current values ​​and trends of the four parameters: temperature, humidity, light intensity, and carbon dioxide concentration. The calculation results show that the D-4 region has the highest similarity to the root node region, reaching 0.82, because its temperature change trend is highly consistent with that of the B-2 region; the F-1 region has a similarity of 0.76, mainly based on the matching of humidity change patterns; although the E-3 region has a low degree of anomaly, it shows a similarity of 0.71.

[0073] The module identifies region D-4 as a primary recommendation node, establishing a priority control association with the root node region. Based on the primary recommendation node, the system calculates the similarity between the remaining regions and region D-4, finding that region F-1 has a similarity of 0.79, primarily based on the synchronous change characteristics of carbon dioxide concentration, thus classifying it as a secondary recommendation node. Continuing the recursive calculation, region G-5 shows a similarity of 0.68 with region F-1, becoming a tertiary recommendation node. As the calculation continues downwards, the similarity between the next candidate region and the current node drops to 0.45, below the system's preset association threshold of 0.5, completing the recommendation chain construction. The covariance correction module then runs independently. This module analyzes the real-time data stream of environmental parameters, finding significant fluctuations in the first-order difference absolute value sequence of the temperature parameter over the past 10 minutes, with a maximum fluctuation intensity reaching 2.3℃ / minute. The system statistically analyzes user command response delay data, finding that the average execution delay of the most recent 5 control commands is 3.2 seconds. The module calculates a correction weight value of 7.36, subsequently using this weight value to dynamically update the noise covariance matrix. The updated matrix is ​​immediately invoked by the recommendation chain building module to adjust the uncertainty factor in similarity calculation. The two modules synchronize data through a shared memory area. The recommendation chain building module reads the updated covariance matrix every 30 seconds and recalculates the similarity relationships between nodes. In this example, the matrix update causes the similarity between region D-4 and the root node to be slightly adjusted from 0.82 to 0.81, and the similarity between region F-1 and region D-4 to be adjusted from 0.79 to 0.78, but the recommendation chain structure remains unchanged.

[0074] The system visualizes the completed multi-layered recommendation chain as a tree-like topology: the root node is region B-2, the first-level child node is region D-4, the second-level child node is region F-1, and the third-level child node is region G-5. Each node includes detailed environmental parameter data and similarity values, along with the last update timestamp. The entire recommendation chain construction process takes 1.8 seconds, and the covariance matrix update takes 0.3 seconds, with real-time performance indicators showing processing latency within acceptable limits. Operators can view the complete recommendation chain structure and understand the relationships between regions through the control interface. The system provides interactive functions, allowing manual adjustment of recommendation chain priorities or the addition of special relationships. All operation records and the system's automatic decision-making process are fully recorded in the audit log, including specific parameters for each similarity calculation, the history of covariance matrix updates, and records of recommendation chain structure adjustments. The collaborative work of the recommendation chain construction and covariance correction modules demonstrates the system's multi-level, adaptive decision-making capabilities in handling complex environmental anomalies. By dynamically adjusting similarity calculation parameters and updating the system noise model in real time, the accuracy and reliability of recommendation chain construction are ensured.

[0075] Example 4: Discovering a potential link between abnormal environmental physical parameters and abnormal user operation commands; Consider a specific application scenario: between 10:00 and 10:30 on May 15, 2023, the system detected a series of abnormal conditions within the isolation shed. The cross-layer correlation module first obtains physical layer data from the abnormal interaction detection module, extracting abnormal fluctuation markers of environmental parameters in all control sub-regions during this period. These markers include key information such as the start time, duration, and amplitude of the fluctuations. Simultaneously, the module obtains all user command records from the user behavior analysis module during the same period, filtering out operation commands marked as abnormal, including commands with abrupt changes in command type, abnormal frequency, or parameter settings exceeding the normal range. The module aligns and matches the abnormal fluctuation periods of the physical layer with the occurrence times of abnormal commands in the command layer, establishing a time window mapping relationship. For each detected abnormal environmental fluctuation event, the module searches for corresponding abnormal user command operations within a time tolerance range of ±2 minutes, forming potential correlation pairs. The module calculates the statistical correlation between the fluctuation range of environmental parameters and the number of abnormal commands in each correlation pair, and uses the Pearson correlation coefficient algorithm to evaluate the linear correlation strength between the two. The correlation coefficient range is [-1, 1], with positive values ​​indicating positive correlation and negative values ​​indicating negative correlation. See Table 1 for the results of the abnormal event correlation analysis in each control sub-region during this period.

[0076] Table 1: Record of Correlation Analysis between Abnormal Fluctuations and Command Operations

[0077]

[0078] The module independently calculates the correlation strength index for each control sub-region. The calculation process employs a sliding time window technique, with the window size dynamically adjusted based on the duration of the anomaly. For anomalies with longer durations, the module uses a segmented calculation method, dividing the long time period into multiple sub-windows to calculate the correlation strength separately, and finally taking the weighted average as the final result. All calculation results are updated in real-time to the correlation knowledge base, which uses a graph structure to store the correlation relationships between regions. Nodes represent anomalies, and edges represent correlation strengths. During module implementation, a distributed computing framework is used to handle large-scale data streams. Physical layer data and instruction layer data are synchronized in real-time through a message middleware to ensure accurate time alignment. The data persistence layer uses a time-series database to store original anomaly records and a relational database to store correlation analysis results, supporting complex queries and historical data analysis. The system provides a visualization interface to display cross-layer correlation analysis results, allowing operators to intuitively view the correlation network and strength distribution between anomalies.

[0079] The identification of abnormal association patterns utilizes machine learning algorithms. The system periodically analyzes and trains historical association data to optimize the association strength calculation model. Training features include multi-dimensional features such as abnormal fluctuation amplitude, instruction type, time interval, and historical association patterns. The model output is used to automatically adjust the association strength threshold, improving the accuracy of anomaly detection. The system also establishes a feedback mechanism; when operators confirm or correct association results, the system automatically updates the training samples, continuously optimizing recognition performance. The module is deployed independently using a microservice architecture and interacts with other modules through standard interfaces. The data processing flow includes multiple stages such as data cleaning, time alignment, feature extraction, association calculation, and result storage. Each stage has quality checkpoints to ensure the reliability of data analysis. The system continuously monitors performance indicators during operation, including processing latency, calculation accuracy, and resource utilization, ensuring stable operation in a real production environment.

[0080] Example 5: Based on the output of the cross-layer correlation module, this module traces the root cause of identified high-correlation-strength anomaly events and generates a reverse control command sequence. Assume that in a certain operational instance, the system detected an anomaly with a cross-layer correlation strength of 0.92 at 14:30 on June 10, 2023. This event involves abnormal fluctuations in temperature and humidity parameters of the control subunit in area C. The tracing execution module first obtains the complete data packet of the anomaly event, including the correlation strength index calculated by the cross-layer correlation module, the anomaly fluctuation period marker, and related user command records. The module then calls the historical database of the environmental data acquisition module to trace the environmental parameter change trajectory of area C within 120 minutes before the anomaly occurred. This data is recorded at a sampling frequency of once per second, forming a high-precision time series.

[0081] The module employs a change point detection algorithm to analyze the trajectory of environmental parameter changes. The algorithm scans time-series data for four parameters: temperature, humidity, light intensity, and carbon dioxide concentration, identifying the moment when a significant deviation from the dynamic environmental parameter baseline first occurs. The scanning process uses a multi-scale sliding window technique, with window sizes ranging from 1 minute to 30 minutes to accommodate the varying characteristics of different parameters. For each detected potential deviation point, the module calculates the magnitude and duration of the deviation from the dynamic baseline, and after comprehensive evaluation, determines the most likely time of occurrence of the root cause. In this example, the system identified the first negative deviation of 0.5 degrees Celsius in the temperature parameter at 14:05:18, and the deviation continued to widen.

[0082] Based on the determined moment of the root cause, the module extracts the environmental data sequence within 10 minutes before and after that moment, calculating the gradient direction of each parameter. The gradient calculation uses the first-order difference method to analyze the continuity and trend characteristics of parameter changes. In this case, the temperature parameter shows a continuously decreasing gradient, the humidity parameter shows a fluctuating upward trend, the light intensity remains relatively stable, and the carbon dioxide concentration increases slightly. Based on the direction of the parameter gradient changes, the module generates a sequence of reverse control commands: for the continuously decreasing temperature parameter, the system generates progressively increasing heating control commands. The first command increases the set temperature by 0.5 degrees Celsius, the second command by 0.8 degrees Celsius, and the third command by 1.0 degrees Celsius, with the magnitude of each command linearly proportional to the initial deviation by 0.8. For the fluctuating upward humidity parameter, progressively decreasing dehumidification commands are generated, with the command magnitude calculated based on the humidity deviation using a proportionality coefficient of 0.6. The control commands for light intensity and carbon dioxide concentration are adjusted appropriately according to their changes, maintaining a relatively conservative control strategy.

[0083] The instruction sequence employs a progressive execution strategy. The first-level instruction is executed immediately after the root cause is identified; the second-level instruction is evaluated after 5 minutes to determine whether to execute; and the third-level instruction, as a backup plan, is activated after the anomaly persists for 10 minutes. After each level of instruction is executed, the system monitors the response of environmental parameters and dynamically adjusts the magnitude and timing of subsequent instructions based on actual feedback. Throughout the process, the module maintains an interactive channel with the user, sending an anomaly root cause analysis report and proposed control instruction plans via terminal devices, awaiting user confirmation or modification before execution. All decision-making processes and execution results are meticulously recorded in the audit log, including the basis for root cause determination, instruction generation logic, and actual execution effects. This data is used for subsequent system optimization and model training. The system implementation employs a redundant design; when the main analysis module fails, the backup module immediately takes over the processing flow, ensuring the continuity of anomaly event handling. Data storage uses a multi-copy mechanism to ensure the integrity and traceability of historical records and audit logs. The entire processing flow can typically be completed within 2 minutes, meeting the real-time requirements of greenhouse environment control.

[0084] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0085] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A control system for a chili pepper breeding isolation shed based on the Internet of Things, characterized in that, The method comprises the following steps: An environmental data acquisition module is used to acquire temperature data, humidity data, light intensity data and carbon dioxide concentration data in the isolation shed in real time; A user behavior analysis module is used to receive and analyze historical data of the user's inputted regulation and control instructions through the terminal device, extract instruction frequency, instruction type and instruction time interval; An isolation shed partition module is used to divide the isolation shed into multiple regulation and control sub-regions according to the data difference obtained by the environmental data acquisition module; A dynamic baseline generation module is used to generate a dynamic environmental parameter baseline for each regulation and control sub-region based on the historical data of the environmental data acquisition module and the instruction data of the user behavior analysis module; An abnormal interaction detection module is used to compare the deviation degree of the real-time environmental data and the dynamic environmental parameter baseline, identify abnormal environmental fluctuations and abnormal user instructions; A root node decision module is used to select the regulation and control sub-region with the most significant environmental fluctuation as the root node region according to the output of the abnormal interaction detection module; A recommendation chain construction module is used to construct a multi-level environmental regulation and control recommendation chain with the root node region as the center and in combination with the real-time environmental data similarity of other regulation and control sub-regions; The recommendation chain construction module specifically comprises the following steps: Take the regulation and control sub-region with the highest environmental parameter similarity as the first-level recommended node; Select the regulation and control sub-region with the second highest similarity as the second-level recommended node based on the first-level recommended node; Recursively construct a multi-level recommended chain until all candidate nodes are covered or the similarity is lower than a preset correlation threshold; The root node and the first-level recommended node have a priority regulation and control correlation, which means that the two regions have the same problem root or the characteristic of coordinated regulation and control.

2. The pepper breeding isolation shed control system based on the Internet of Things according to claim 1, characterized in that, The user behavior analysis module specifically comprises the following steps: Divide the user instruction historical data into stable instruction periods and changing instruction periods according to the time sequence; Extract the average execution frequency of each instruction type in the stable instruction period and the time and direction of instruction type mutation in the changing instruction period; Select a set of instruction types to be analyzed according to the instruction frequency difference coefficient of the stable period and the changing period.

3. The pepper breeding isolation shed control system based on the Internet of Things according to claim 2, characterized in that, The dynamic baseline generation module specifically comprises the following steps: For each regulation and control sub-region, calculate the ratio of the standard deviation of the environmental parameter in the stable instruction period to the standard deviation of the environmental parameter in the changing instruction period as a baseline dynamic adjustment factor; Take the average value of the environmental parameter in a continuous preset period before the user instruction type mutation time as the temporary environmental baseline of the regulation and control sub-region; Weight and correct the temporary environmental baseline according to the baseline dynamic adjustment factor to generate the final dynamic environmental parameter baseline.

4. The pepper breeding isolation shed control system based on the Internet of Things according to claim 3, characterized in that, The abnormal interaction detection module specifically comprises the following steps: For each regulation and control sub-region, calculate the Euclidean distance between the current environmental parameter and the dynamic environmental parameter baseline in real time; When the Euclidean distance exceeds a preset threshold, extract the user instruction type and execution timestamp corresponding to the time; If the interval between the user instruction execution timestamp and the starting time of the environmental parameter fluctuation exceeds a preset delay threshold, mark it as an abnormal interaction event.

5. The pepper breeding isolation shed control system based on the Internet of Things according to claim 4, characterized in that, The root node decision module specifically comprises the following steps: Calculate the maximum value of the Euclidean distance in the abnormal interaction events of all regulation and control sub-regions; The maximum value corresponds to the regulatory sub-region as the root node region, and the remaining regions are arranged in descending order of Euclidean distance as candidate nodes; Record the similarity matrix of the root node region and each candidate node region at the abnormal fluctuation time of the environmental parameter.

6. The pepper breeding isolation shelter control system based on the Internet of Things according to claim 1, characterized in that, The system also includes a covariance correction module, which dynamically corrects the noise covariance matrix of the environmental regulation instruction based on the real-time data fluctuation amplitude of the environmental data acquisition module and the response delay of the user instruction, specifically including: Calculate the first-order difference absolute value sequence of the real-time sampling sequence of the environmental parameter, and take the maximum value in the sequence as the fluctuation intensity reference; Statistical user instruction from the generation to the execution of the delay time, calculate the product of the delay time and the fluctuation intensity reference as the correction weight; Use the correction weight to dynamically update the initial value of the noise covariance matrix.

7. The pepper breeding isolation shed control system based on the Internet of Things according to claim 6, characterized in that, The system also includes a cross-layer association module that integrates the physical layer data of the environmental data acquisition module and the instruction layer data of the user behavior analysis module, identifies cross-layer abnormal association patterns, and specifically includes: Extract the time period marker of the abnormal fluctuation of the physical layer environmental parameter; Retrieve all abnormal instruction records in the user instruction layer within the same period; Calculate the Pearson correlation coefficient of the environmental parameter fluctuation amplitude and the number of abnormal instructions as the cross-layer association strength indicator.

8. The pepper breeding isolation shed control system based on the Internet of Things according to claim 7, characterized in that, The system also includes a trace execution module that traces the root cause of abnormal environmental fluctuations and generates a targeted regulation instruction sequence based on the output of the cross-layer association module, specifically including: For abnormal events with cross-layer association strength exceeding the preset threshold, backtrack the environmental parameter change trajectory of the root node region before the abnormal fluctuation; match the time when the first deviation from the dynamic environmental parameter baseline occurs in the trajectory as the root cause occurrence time; based on the gradient change direction of the environmental data collected before and after the root cause occurrence time, generate a reverse regulation instruction sequence.

9. The pepper breeding isolation shed control system based on the Internet of Things according to claim 8, characterized in that, The generation logic of the reverse regulation instruction sequence includes: If the environmental parameter gradient change direction is continuously rising, generate a step-by-step decreasing regulation instruction; If the environmental parameter gradient change direction is continuously falling, generate a step-by-step increasing regulation instruction; The amplitude of each instruction is in a linear proportional relationship with the environmental parameter deviation amplitude at the root cause occurrence time.

Images