A chicken farm environment monitoring system based on intelligent sensors

Abstract

The present application relates to the field of intelligent breeding Internet of Things technology, in particular to a chicken farm environment monitoring system based on intelligent sensors, comprising: data acquisition, anomaly detection, deep diagnosis and strategy generation modules. By laying multiple sensors in the chicken farm, time-aligned environmental data sequences are obtained. The anomaly detection module fuses multi-source data into a multi-dimensional environmental feature vector and calculates an environmental anomaly index to comprehensively judge whether the state deviates from the normal. When an anomaly is determined, the deep diagnosis module is triggered to perform pattern mining on the anomaly period data, identify the core environmental factors and their evolution patterns, and generate a diagnosis report. The strategy generation module matches the strategy library according to the report to generate a targeted regulation instruction sequence. This method realizes accurate evaluation of the overall state of the breeding environment and intelligent diagnosis of the root cause of the anomaly, improving the automation and regulation accuracy of environmental management.

Description

Technical Field

[0001] This invention relates to the field of smart farming Internet of Things technology, and in particular to a chicken farm environmental monitoring system based on smart sensors. Background Technology

[0002] Traditional chicken farm environmental monitoring relies on independent sensors deployed scattered throughout the chicken houses, and generally employs single-parameter alarm mechanisms based on fixed thresholds. This method can only reflect the instantaneous state of a single environmental parameter and cannot comprehensively assess the overall environmental state resulting from the coupling of multiple factors such as temperature, humidity, and gas concentration. Due to the complex correlations between environmental parameters, this isolated and static threshold judgment method is prone to false alarms due to local fluctuations and struggles to capture potential abnormal changes occurring simultaneously with multiple parameters, leading to blind spots and delays in assessing overall environmental deterioration.

[0003] After an alarm is triggered, existing solutions typically lack automated, in-depth analysis capabilities. Managers must rely on personal experience to manually backtrack and compare historical data from various sensors during the abnormal period to attempt to deduce the root cause. This process is inefficient, and the accuracy and consistency of diagnostic results are difficult to guarantee. The system's functionality stops at "status alarms," ​​failing to achieve "root cause diagnosis," creating a gap in decision-making information between problem discovery and implementation of control measures, and failing to provide direct evidence for precise environmental intervention.

[0004] A technical solution is needed to assess the overall condition of the aquaculture environment and diagnose anomalies in real time. This solution must overcome the limitations of single-parameter threshold methods, achieving comprehensive condition judgment based on multi-source data fusion; and after identifying anomalies, it should automatically analyze the causes and evolution patterns of the anomalies, transforming alarm signals into diagnostic conclusions that can guide operations. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a chicken farm environmental monitoring system based on intelligent sensors.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a chicken farm environmental monitoring system based on intelligent sensors, comprising:

[0007] The data acquisition and cleaning module deploys multiple smart sensor nodes at predetermined locations in the chicken farm to acquire time-aligned environmental data sequences from the chicken farm.

[0008] The environmental anomaly detection module constructs an environmental state feature extraction model, calculates a multi-dimensional environmental feature vector representing the overall environmental state of the chicken farm based on the time-aligned environmental data sequence, establishes an environmental anomaly detection model, uses the multi-dimensional environmental feature vector to calculate the environmental anomaly index at each sampling time, and determines whether the current environmental state deviates from the normal range based on the environmental anomaly index.

[0009] The deep analysis and diagnosis module, when the environmental anomaly detection model determines that the current environmental state deviates from the normal range, triggers the environmental data analysis process, performs deep pattern mining on the environmental data during the deviation period, identifies the core environmental factors that cause the environmental state deviation and their abnormal evolution patterns through the environmental data analysis process, and generates an environmental anomaly diagnosis report.

[0010] The control strategy generation module generates a targeted environmental control instruction sequence by matching the environmental anomaly diagnosis report with a preset environmental control strategy library.

[0011] As a further aspect of the present invention, the step of deploying multiple intelligent sensor nodes at predetermined locations in the chicken farm to acquire time-aligned environmental data sequences of the chicken farm specifically includes:

[0012] Multiple smart sensor nodes are deployed at predetermined locations in the chicken farm to acquire raw environmental data. The raw environmental data includes at least temperature data, humidity data, gas concentration data, and light intensity data. The raw environmental data is cleaned to filter out abnormal data points caused by sensor transient failures or external instantaneous interference, resulting in cleaned environmental data.

[0013] The cleaned environmental data is time-synchronized and aligned so that data of the same type from different smart sensor nodes remain aligned on the time axis, forming a time-aligned environmental data sequence.

[0014] The acquisition of raw environmental data includes:

[0015] Based on the distribution of chicken activity areas, ventilation paths, and layout of environmental control equipment within the chicken farm, the deployment location and density of intelligent sensor nodes are planned.

[0016] Install intelligent sensor nodes at each deployment point, which integrate temperature sensing units, humidity sensing units, gas sensing units, and light sensing units.

[0017] Set a uniform initial data acquisition frequency and start all smart sensor nodes to periodically collect environmental parameters;

[0018] Each smart sensor node locally encapsulates the collected temperature, humidity, gas concentration, and light intensity data, adding node identifiers, collection timestamps, and location information to form raw environmental data.

[0019] The raw environmental data is transmitted to the central data processing platform via a wireless sensor network.

[0020] As a further aspect of the present invention, the step of cleaning the raw environmental data to filter out abnormal data points caused by sensor transient failures or external transient interference, thereby obtaining cleaned environmental data, specifically includes:

[0021] The system receives the raw environmental data and parses it to obtain temperature data, humidity data, gas concentration data, light intensity data, and corresponding node identifiers, timestamps, and location information.

[0022] For each environmental parameter data of each smart sensor node, the mean and standard deviation of statistical data based on a sliding time window are used;

[0023] The current data point is compared with the mean of the historical data within the sliding time window. If the difference exceeds the standard deviation of a preset multiple, the current data point is determined to be an abnormal data point.

[0024] Replace the values ​​of the data points marked as outliers with the mean of the historical data within the sliding time window or the values ​​generated by the interpolation algorithm.

[0025] After anomaly detection and replacement of all environmental parameter data, the cleaned environmental data is output.

[0026] As a further aspect of the present invention, the step of time-synchronizing and aligning the cleaned environmental data, so that data of the same type from different smart sensor nodes remain aligned on the time axis to form a time-aligned environmental data sequence, specifically includes:

[0027] Using the system clock of the central data processing platform as a reference, calibrate the acquisition timestamp in the data packet of each smart sensor node;

[0028] Set a uniform time alignment granularity to divide the timeline into continuous equal-length slices;

[0029] For each time slice, iterate through all the cleaned environmental data reported by all smart sensor nodes within this time slice;

[0030] If a smart sensor node has multiple data points within a certain time slice, the average value of the environmental parameter data of the smart sensor node within the time slice is taken as the representative value of the smart sensor node within this time slice.

[0031] If a smart sensor node has no data points in a certain time slice, the interpolation of the data from adjacent time slices of the smart sensor node is used as the representative value of the time slice.

[0032] The representative values ​​of various environmental parameters of all smart sensor nodes in the same time slice are arranged and combined to generate an environmental state snapshot of the time slice. The environmental state snapshots of all time slices are arranged in chronological order to form a time-aligned environmental data sequence.

[0033] As a further aspect of the present invention, the construction of the environmental state feature extraction model, which calculates a multidimensional environmental feature vector representing the overall environmental state of the chicken farm based on the time-aligned environmental data sequence, specifically includes:

[0034] Define the dimensions of the multidimensional environmental feature vector, including spatial distribution feature dimension, temporal variation feature dimension, and parameter coupling feature dimension;

[0035] Calculate spatial distribution characteristics: Select the same environmental parameter data of all smart sensor nodes in the same time slice, and calculate its global average, maximum, minimum and spatial distribution variance;

[0036] Calculate the time variation characteristics: For each environmental parameter data of each smart sensor node, extract its value in multiple consecutive time slices, and calculate its short-term trend slope and fluctuation intensity.

[0037] Calculate parameter coupling characteristics: Analyze the correlation coefficients between different environmental parameters on the same smart sensor node at the same time slice, such as the correlation between temperature and humidity;

[0038] The spatial distribution features, temporal variation features, and parameter coupling features calculated for each time slice are concatenated to form a multidimensional environmental feature vector corresponding to the time slice.

[0039] As a further aspect of the present invention, the step of establishing an environmental anomaly detection model, calculating an environmental anomaly index at each sampling time using the multi-dimensional environmental feature vector, and determining whether the current environmental state deviates from the normal range based on the environmental anomaly index, specifically includes:

[0040] Collect multidimensional environmental feature vectors from the history of chicken farms during normal production cycles to form a normal environmental feature sample set;

[0041] Based on the normal environment feature sample set, a single-class classification model is trained, which can describe the distribution range of normal environment features in the feature space.

[0042] For the multidimensional environmental feature vector of the current time slice obtained in real time calculation, input it into the single-class classification model to calculate its distance from the center of normal distribution or the probability of belonging to normal distribution;

[0043] The distance or probability is converted into an environmental anomaly index. The higher the environmental anomaly index, the greater the degree of deviation from the normal state.

[0044] A threshold for the environmental anomaly index is set. When the calculated environmental anomaly index exceeds the threshold, the current environmental state is determined to deviate from the normal range.

[0045] As a further aspect of the present invention, when the environmental anomaly detection model determines that the current environmental state deviates from the normal range, an environmental data analysis process is triggered to perform deep pattern mining on the environmental data during the deviation period, specifically including:

[0046] Record the start and end times of environmental conditions deviating from the normal range to determine the period of deviation;

[0047] Extract time-aligned environmental data sequences and corresponding multi-dimensional environmental feature vectors within the off-time period;

[0048] Cluster analysis is performed on the environmental data during the off-period time period to identify a group of smart sensor nodes that exhibit similar change patterns during the off-period time period;

[0049] Principal component analysis was performed on the multidimensional environmental feature vectors within the deviation period to identify the main feature dimensions that led to the increase in the environmental anomaly index.

[0050] The results of the cluster analysis are correlated with the results of the principal component analysis to analyze how changes in the feature dimensions of a specific group of nodes collectively lead to anomalies in the overall environment.

[0051] As a further aspect of the present invention, the step of identifying the core environmental factors causing deviations in environmental status and their abnormal evolution patterns through the environmental data analysis process, and generating an environmental anomaly diagnosis report, specifically includes:

[0052] Based on the correlation results of the cluster analysis and principal component analysis, the environmental parameters that changed most significantly during the deviation period were identified, and these environmental parameters are the core environmental factors.

[0053] Track the time series changes of the core environmental factors during the deviation period and extract their change pattern characteristics, including the time of abrupt change, the rate of change, and the duration.

[0054] Analyze the interaction between the core environmental factors and other environmental parameters during the deviation period to identify abnormal coupling evolution patterns;

[0055] The identification results of the core environmental factors, the description of their abnormal evolution patterns, and the analysis of their linkage with other parameters are integrated to generate a structured environmental anomaly diagnostic report.

[0056] As a further aspect of the present invention, the step of generating a targeted environmental control instruction sequence by matching the environmental anomaly diagnosis report with a preset environmental control strategy library specifically includes:

[0057] Maintain an environmental control strategy library, which stores the mapping relationship between different environmental anomaly patterns and corresponding control measures;

[0058] The core environmental factors and their abnormal evolution patterns described in the environmental anomaly diagnosis report are matched with the abnormal pattern entries in the environmental regulation strategy library.

[0059] Find the abnormal pattern entries with the highest similarity and obtain the set of predefined environmental control measures associated with them;

[0060] Based on the specific parameter changes in the environmental anomaly diagnosis report, the parameters in the predefined set of environmental control measures are instantiated, such as setting the target speed of the fan or the target opening of the ventilation window.

[0061] The instantiated control measures are arranged according to the preset execution logic and timing to generate a specific executable environmental control instruction sequence.

[0062] As a further aspect of the present invention, the system further includes:

[0063] The execution feedback evaluation module sends the environmental control command sequence to the environmental control execution equipment within the chicken farm, continuously receives feedback status data from the environmental control execution equipment, and evaluates and dynamically adjusts the execution effect of the environmental control command sequence in conjunction with the latest raw environmental data. Specifically, this includes:

[0064] The environmental control command sequence is sent to the corresponding environmental control execution device through the control network;

[0065] After executing the command, the environmental control equipment sends back its current working status as feedback status data;

[0066] During the execution of the environmental control command sequence, new raw environmental data continues to be collected and processed through cleaning, alignment, feature extraction, and anomaly index calculation.

[0067] Monitor the changing trends of environmental anomaly indices calculated using new raw environmental data, and evaluate the effectiveness of control measures;

[0068] If the environmental anomaly index does not fall back below the normal threshold within the expected time, or if the feedback status data shows that the equipment has not executed the instructions, the environmental control strategy will be re-matched or corrected based on the latest multidimensional environmental feature vector and environmental anomaly diagnosis report, and a new environmental control instruction sequence will be generated and issued.

[0069] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0070] By constructing a feature extraction model, time-series data collected by multiple sensors are fused and calculated into a multi-dimensional feature vector representing the overall environmental state. Based on this, a model is established to calculate a dynamic environmental anomaly index. This achieves a shift from relying on single-parameter static thresholds to multi-dimensional, systematic quantitative assessment of the environment. The system can continuously learn normal patterns and adaptively identify abnormal states that deviate from the overall pattern. Even if a single parameter does not exceed the limit, it can effectively detect comprehensive anomalies caused by imbalances in the correlation of multiple parameters, reducing misjudgments caused by local random fluctuations and improving the sensitivity to complex potential anomalies.

[0071] When an anomaly index indicates an environmental deviation, the system automatically initiates a deep analysis process of high-dimensional time-series data for the abnormal period. This process employs time-series pattern mining and feature attribution analysis techniques to automatically identify the core driving factors contributing most to the current abnormal state from numerous environmental parameters, and analyzes the dynamic changes and interactions of these factors before and after the anomaly. This process transforms a single alarm signal into structured diagnostic conclusions, clearly revealing the main causes and development process of the abnormal event.

[0072] Based on the diagnostic results, the system can match a pre-set environmental control strategy library to generate a targeted sequence of control instructions. This allows environmental intervention measures to be directly derived from the anomaly diagnosis results, avoiding the lag and uncertainty caused by relying on trial and error based on human experience. The entire process achieves fully automated decision support across the entire chain, from comprehensive state perception and intelligent anomaly tracing to precise strategy generation. Attached Figure Description

[0073] Figure 1 This is a timing diagram of the chicken farm environmental monitoring system based on intelligent sensors described in this invention;

[0074] Figure 2 A flowchart for cleaning raw environmental data;

[0075] Figure 3 A flowchart for calculating multidimensional environmental feature vectors. Detailed Implementation

[0076] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0077] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0078] See Figure 1 The data acquisition and cleaning module deploys multiple intelligent sensor nodes at predetermined locations within the chicken farm to acquire time-aligned environmental data sequences. The environmental anomaly detection module constructs an environmental state feature extraction model, calculates a multi-dimensional environmental feature vector representing the overall environmental state of the chicken farm based on the time-aligned environmental data sequences, establishes an environmental anomaly detection model, calculates an environmental anomaly index for each sampling moment using the multi-dimensional environmental feature vector, and determines whether the current environmental state deviates from the normal range based on the environmental anomaly index. The deep analysis and diagnosis module triggers an environmental data analysis process when the environmental anomaly detection model determines that the current environmental state deviates from the normal range. It performs deep pattern mining on the environmental data during the deviation period, identifies the core environmental factors causing the environmental state deviation and their abnormal evolution patterns through the environmental data analysis process, and generates an environmental anomaly diagnosis report. The control strategy generation module matches the environmental anomaly diagnosis report with a preset environmental control strategy library to generate a targeted environmental control instruction sequence.

[0079] See Figure 2In one embodiment of the present invention, the deployment location and density of intelligent sensor nodes are planned according to the distribution of the chicken flock's activity areas, ventilation paths, and the layout of environmental control equipment in the chicken farm. The deployment locations include the four corners of the chicken house, the central area, and areas near the air inlets and outlets. At each planned deployment point, an intelligent sensor node integrating a temperature sensing unit, a humidity sensing unit, a gas sensing unit, and a light sensing unit is installed. The gas sensing unit is used to monitor ammonia and carbon dioxide concentrations. A uniform initial data acquisition frequency of once per minute is set, and all intelligent sensor nodes are started to periodically collect environmental parameters. Each intelligent sensor node locally encapsulates the collected temperature data, humidity data, gas concentration data, and light intensity data, adding a node identifier, acquisition timestamp, and location information to form a raw environmental data packet. The raw environmental data packet is transmitted to the central data processing platform deployed in the monitoring room via a LoRa-based wireless sensor network.

[0080] In practice, the raw environmental data undergoes data cleaning to filter out abnormal data points caused by sensor transient failures or transient external interference. The central data processing platform receives the raw environmental data packets and parses them to obtain temperature data, humidity data, gas concentration data, light intensity data, and corresponding node identifiers, timestamps, and location information. For each environmental parameter data sequence of each smart sensor node, based on the mean and standard deviation of the statistical data within a sliding time window, the sliding time window covers the most recent continuous... 10 historical data points. Compare the current data point with the mean of the historical data within the sliding time window. If the absolute value of the difference exceeds 100, 10 data points are considered. If the standard deviation is more than one time the current data point, it is considered an outlier. The values ​​marked as outliers are replaced with the mean of historical data within the sliding time window or values ​​generated by linear interpolation. After outlier detection and replacement for all environmental parameter data, the cleaned environmental data is output. The formula for determining outliers is as follows:

[0081]

[0082] in: This represents the value of the current data point. This represents the arithmetic mean of historical data within the sliding time window. This represents the standard deviation of historical data within the sliding time window. It is a preset constant coefficient that is greater than zero.

[0083] In practice, the cleaned environmental data is time-synchronized and aligned to ensure that data of the same type from different smart sensor nodes are aligned on the timeline, forming a time-aligned environmental data sequence. The acquisition timestamp in the data packet of each smart sensor node is calibrated based on the system clock of the central data processing platform to correct time deviations caused by device clock drift or network latency. A uniform time alignment granularity of five minutes is set, dividing the timeline into consecutive equal-length time slices. For each time slice, all cleaned environmental data reported by all smart sensor nodes within that time slice are traversed. If a smart sensor node has multiple data points within a time slice, the average value of the environmental parameter data of the smart sensor nodes within that time slice is taken as the representative value of the smart sensor node within that time slice. If a smart sensor node has no data points within a time slice, the linear interpolation of the data from adjacent time slices is used as the representative value of that time slice. The representative values ​​of various environmental parameters of all smart sensor nodes in the same time slice are arranged and combined to generate an environmental state snapshot of the time slice. All environmental state snapshots of the time slices are arranged in chronological order to form the final time-aligned environmental data sequence.

[0084] See Figure 3 In one embodiment of the present invention, the dimensions of the multidimensional environmental feature vector include spatial distribution feature dimension, temporal variation feature dimension, and parameter coupling feature dimension. Taking a chicken coop with six smart sensor nodes as an example, the time-aligned environmental data sequence is divided into five-minute time slices. Each time slice contains the cleaned values ​​of four parameters reported by all nodes: temperature, humidity, ammonia concentration, and light intensity. For a specific time slice, when calculating the spatial distribution feature, the same environmental parameter data, such as temperature data, is selected from all smart sensor nodes within that time slice. The values ​​of the six nodes are 22.1°C, 23.5°C, 25.3°C, 22.8°C, 24.1°C, and 21.9°C, respectively. The global average of this set of data is calculated to be 23.28°C, the maximum value is 25.3°C, and the minimum value is 21.9°C. Spatial distribution variance. The formula used to quantify the spatial dispersion of data is as follows:

[0085]

[0086] in: Representing the The parameter values ​​of each smart sensor node, This represents the arithmetic mean of the values ​​of this parameter across all nodes. This represents the total number of smart sensor nodes. Substituting the six temperature values ​​into the formula, the spatial distribution variance is calculated to be approximately 1.31. This calculation process is repeated for humidity, ammonia concentration, and light intensity parameters to obtain their respective spatial distribution characteristic values.

[0087] In some embodiments, time-varying features are calculated for each environmental parameter data for each smart sensor node. A smart sensor node is selected, and its temperature value sequence over five consecutive time slices is extracted as 22.0°C, 22.3°C, 22.8°C, 23.5°C, and 24.1°C. The short-term trend slope is obtained by performing a first-order linear fit on this sequence, indicating that the temperature rises at a rate of approximately 0.0007°C per second within this time window. The fluctuation intensity is quantified by calculating the standard deviation of the sequence, which is approximately 0.77°C. The same calculation is performed for each environmental parameter of each smart sensor node in the chicken coop, thereby generating a set of time-varying features for each node-parameter pair. It can be understood that the time-varying features reflect the dynamic behavior of the environmental parameters over time.

[0088] Optionally, the correlation between different environmental parameters on the same time slice and the same smart sensor node can be analyzed using parameter coupling feature analysis. For time slice t and smart sensor node A, the temperature and humidity values ​​of the node in that time slice are extracted, assumed to be 23.5°C and 65%, respectively. By analyzing the temperature and humidity data sequences of node A in multiple consecutive time slices, the Pearson correlation coefficient between the two is calculated. If the correlation coefficient is -0.85, it indicates a strong negative correlation between temperature and humidity in the current region. This calculation process requires traversing all smart sensor nodes and calculating the correlation coefficients of key parameter pairs (such as temperature-humidity, temperature-ammonia concentration), with the correlation coefficient ranging from -1 to 1. In some embodiments, all spatial distribution features, temporal variation features, and parameter coupling features calculated for each time slice are sequentially concatenated to form a fixed-length multidimensional environmental feature vector corresponding to that time slice. It can be understood that this multidimensional environmental feature vector integrates multidimensional information about the relationship between environmental states in space, time, and different parameters.

[0089] In one embodiment of the present invention, establishing an environmental anomaly detection model requires collecting historical multidimensional environmental feature vectors from a chicken farm during its normal production cycle. A normal production cycle is defined as several consecutive working days during which the flock is healthy and the environmental control equipment operates stably. The system collects multidimensional environmental feature vectors corresponding to all time slices within this period, forming a normal environmental feature sample set. A single-class classification model is trained based on this normal environmental feature sample set. The single-class classification model employs a support vector machine (SVM) algorithm. This algorithm describes the distribution range of normal environmental features in the feature space by finding a hypersphere with the smallest volume in the high-dimensional feature space; samples within the hypersphere are considered normal. For the multidimensional environmental feature vector of the current time slice calculated in real time, it is input into the trained single-class classification model, and the distance from this multidimensional environmental feature vector to the center of the hypersphere of the SVM is calculated. The calculated distance value is converted into an environmental anomaly index by dividing the distance value by the hypersphere radius and then linearly scaling it. A higher environmental anomaly index indicates a greater deviation from the normal state. The threshold for the environmental anomaly index is set at 1.5. When the calculated environmental anomaly index exceeds the threshold of 1.5, the system determines that the current environmental state deviates from the normal range. This can be understood as a support vector machine model being trained using only normal historical data, without the need for abnormal samples.

[0090] In some embodiments, the environmental anomaly detection model automatically triggers the environmental data analysis process when it determines that the current environmental state deviates from the normal range. The system records the start time of the environmental state deviation from the normal range as 08:30 and the end time as 10:15, thus determining the deviation period as 08:30 to 10:15. The system extracts the time-aligned environmental data sequences and corresponding multidimensional environmental feature vectors for all time slices within the deviation period. Cluster analysis is performed on the environmental data during the deviation period. The K-means algorithm based on Euclidean distance is used to analyze the temperature change curves of six smart sensor nodes during the deviation period, identifying a group of smart sensor nodes exhibiting similar change patterns. For example, nodes 1, 2, and 5 are grouped together, and the temperatures of these nodes rise rapidly and synchronously after 08:45. Principal component analysis is performed on the multidimensional environmental feature vectors within the deviation period. Principal component analysis projects high-dimensional features onto a few principal components, identifying the main feature dimensions that cause the increase in the environmental anomaly index. The loading matrix of the first principal component shows that the spatial distribution variance and temporal change slope features related to temperature contribute the most. Correlating the results of cluster analysis with those of principal component analysis reveals that the increased spatial variance of temperature distribution and the enhanced short-term upward trend in temperature within the node cluster consisting of nodes 1, 2, and 5 jointly contributed to the increase in the overall environmental anomaly index. It can be understood that the correlation between cluster analysis and principal component analysis aims to locate the spatial source of the anomaly and explain its driving characteristics.

[0091] Optionally, the calculation of the environmental anomaly index can be expressed by the following formula:

[0092]

[0093] in: This represents the calculated environmental anomaly index. Represents the current time slice The multidimensional environmental feature vector This represents the center vector of the normal distribution in a uniclass classification model. Representative vector With the center vector The Euclidean distance between them The reference radius represents the range of the normal distribution defined by a single-class classification model. It is a preset scaling factor used to adjust the scale of the index.

[0094] In one embodiment of the present invention, the environmental parameter that changes most significantly during the deviation period is determined based on the correlation results of cluster analysis and principal component analysis. During the deviation period from 08:30 to 10:15, principal component analysis shows that the variance contribution rate of the first principal component reaches 85%, and the weight of the feature dimension related to temperature in its loading matrix is ​​much higher than that of other parameters such as humidity and ammonia concentration. Cluster analysis further identifies that the temperature change patterns of the node group composed of nodes 1, 2, and 5 are highly consistent and abnormally significant, thus determining "temperature" as the core environmental factor. Tracking the time series changes of the core environmental factor temperature during the deviation period, its change pattern characteristics are extracted. The temperature data recorded by the system shows that the temperature starts from 23°C at 08:30 and rises continuously, reaching a peak of 28°C at 09:15, and then slowly decreases. The abrupt change occurs at 08:45, and the rate of change is calculated by measuring the temperature change per unit time, with a duration of 105 minutes. Analyzing the linkage between the core environmental factor temperature and other environmental parameters during the deviation period, the correlation coefficient between the temperature and humidity sequences is calculated to be -0.92, identifying a coupled evolution pattern of abnormal temperature rise accompanied by a sharp drop in humidity. The identification results of core environmental factors, the description of their abnormal evolution patterns, and the analysis of their correlation with other parameters are integrated to generate a structured environmental anomaly diagnostic report. The report uses a predefined XML format and includes fields such as event number, time window, core factors, and evolution pattern description. See Table 1.

[0095] Table 1: Evolution Characteristics of Temperature Anomaly Periods

[0096] Time point Temperature value (°C) Calculate the rate of change (°C / minute) 08:30 23.0 0.00 08:45 24.5 0.10 09:00 26.5 0.13 09:15 28.0 0.10 09:30 27.5 -0.03 09:45 26.8 -0.05 10:00 26.0 -0.05 10:15 25.5 -0.03

[0097] In some embodiments, rate of change The calculation formula is as follows:

[0098]

[0099] in: The average rate of change of the representative parameter over the time interval. Representative parameters at the start time The value, The representative parameter at the end time The numerical value. According to the table data, the rate of change at 08:45. The calculation is (24.5-23.0) / (15)=0.10°C / minute.

[0100] In practical implementation, a targeted sequence of environmental control instructions is generated by matching the environmental anomaly diagnosis report with a pre-defined environmental control strategy library. The system maintains an environmental control strategy library, stored in a relational database table format, containing fields for anomaly pattern description, core environmental factors, and a set of control measures. The core environmental factor "temperature" described in the environmental anomaly diagnosis report, along with its anomaly evolution pattern "continuous rapid rise with a strong negative correlation to humidity," is matched against anomaly pattern entries in the environmental control strategy library using a weighted similarity algorithm based on text keywords and numerical features. The anomaly pattern entry with the highest similarity is identified, and its associated predefined set of environmental control measures is obtained. This set includes three measures: "starting the longitudinal ventilation system," "adjusting the evaporative cooling pad pump operating ratio," and "dimming the light intensity." Based on the specific parameter changes in the environmental anomaly diagnosis report, the parameters in the predefined set of environmental control measures are instantiated; for example, the target speed of the fan is set to 70% of its rated speed, and the target opening degree of the ventilation window is set to 60%. The instantiated control measures are arranged according to a preset execution logic and timing sequence to generate a specific executable environmental control instruction sequence. The instruction sequence format is a JSON array, with the first instruction being "{Device:'Wind Turbine Unit 1', Instruction:'set_speed', Parameter:70%, Delay:0 seconds}", and the second instruction being executed after 5 minutes. In some embodiments, the environmental control strategy library is updated through manual configuration or machine learning based on historical control effect data. It can be understood that the environmental control strategy library is the knowledge foundation for achieving automated and precise control.

[0101] In one embodiment of the present invention, an environmental control command sequence is sent to the environmental control execution equipment within the chicken farm. The system sends the generated environmental control command sequence to the corresponding environmental control execution equipment via a preset control network. The environmental control command sequence is encapsulated in JSON format, including equipment identification, command type, target parameters, and execution timing information. After executing the command, the environmental control execution equipment sends back its current operating status as feedback status data; for example, the fan sends back the actual rotational speed as 68% of the rated speed, and the evaporative cooling pad pump sends back the operating current as 1.5 amps. During the execution of the environmental control command sequence, the system continues to collect new raw environmental data from all intelligent sensor nodes at the original acquisition frequency. The new raw environmental data undergoes the same cleaning and time synchronization alignment process as in the embodiment, and then undergoes the same feature extraction process as in the embodiment to obtain a new multidimensional environmental feature vector. Based on the new multidimensional environmental feature vector, the environmental anomaly detection model calculates the latest environmental anomaly index, and monitors the changing trend of the environmental anomaly index calculated under the new raw environmental data to evaluate the effectiveness of the control measures. If the environmental anomaly index does not fall back below the normal threshold within the expected time, or if the feedback status data shows that the equipment has not executed the instructions, then based on the latest multidimensional environmental feature vector and the updated environmental anomaly diagnosis report, the environmental control strategy will be re-matched or corrected, and a new environmental control instruction sequence will be generated and issued.

[0102] In some embodiments, the effectiveness of regulatory measures can be assessed by quantifying the difference between the expected trend and the actual trend. Define the assessment function. To measure the environmental anomaly index During the regulation period The degree to which the actual decline matches the expected decline model. An optional evaluation function is as follows:

[0103]

[0104] in: This represents the evaluation function value, which ranges from 0 to 1. The closer the value is to 1, the better the control effect matches expectations. This represents the number of time slices that have elapsed since the start of the regulation. Representative at the The environmental anomaly index was actually calculated for each time slice. The representative is based on the expected decline model in the first... The expected environmental anomaly index for each time slice. Represents the actual environmental anomaly index sequence exist The average value over several time slices. This evaluation function, based on the concept of goodness of fit, quantifies the degree to which the actual data deviates from the expected model.

[0105] In practice, the dynamic adjustment process is triggered based on the evaluation results. If the evaluation function value... If the environmental anomaly index falls below a preset threshold, or fails to drop below the normal threshold within three time slices after the execution of the control command sequence, the system determines that the current control strategy is ineffective. The system combines the latest multidimensional environmental feature vectors collected over a period of time with the core environmental factor information in the environmental anomaly diagnosis report, and re-matches them with the environmental control strategy library, or modifies the control parameters in the already matched strategies by a certain proportion, thereby generating a new environmental control command sequence. This new environmental control command sequence is then issued again through the control network, initiating a new round of control and evaluation cycles.

[0106] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A chicken farm environmental monitoring system based on intelligent sensors, characterized in that, Includes the following steps: The data acquisition and cleaning module deploys multiple smart sensor nodes at predetermined locations in the chicken farm to acquire time-aligned environmental data sequences from the chicken farm. The environmental anomaly detection module constructs an environmental state feature extraction model, calculates a multi-dimensional environmental feature vector representing the overall environmental state of the chicken farm based on the time-aligned environmental data sequence, establishes an environmental anomaly detection model, uses the multi-dimensional environmental feature vector to calculate the environmental anomaly index at each sampling time, and determines whether the current environmental state deviates from the normal range based on the environmental anomaly index. The deep analysis and diagnosis module, when the environmental anomaly detection model determines that the current environmental state deviates from the normal range, triggers the environmental data analysis process, performs deep pattern mining on the environmental data during the deviation period, identifies the core environmental factors that cause the environmental state deviation and their abnormal evolution patterns through the environmental data analysis process, and generates an environmental anomaly diagnosis report. The control strategy generation module generates a targeted environmental control instruction sequence by matching the environmental anomaly diagnosis report with a preset environmental control strategy library.

2. The chicken farm environmental monitoring system based on intelligent sensors according to claim 1, characterized in that, The deployment of multiple smart sensor nodes at predetermined locations within the chicken farm to acquire time-aligned environmental data sequences specifically includes: Multiple smart sensor nodes are deployed at predetermined locations in the chicken farm to acquire raw environmental data. The raw environmental data includes at least temperature data, humidity data, gas concentration data, and light intensity data. The raw environmental data is cleaned to filter out abnormal data points caused by sensor transient failures or external instantaneous interference, resulting in cleaned environmental data. The cleaned environmental data is time-synchronized and aligned so that data of the same type from different smart sensor nodes remain aligned on the time axis, forming a time-aligned environmental data sequence. The acquisition of raw environmental data includes: Based on the distribution of chicken activity areas, ventilation paths, and layout of environmental control equipment within the chicken farm, the deployment location and density of intelligent sensor nodes are planned. Install intelligent sensor nodes at each deployment point, which integrate temperature sensing units, humidity sensing units, gas sensing units, and light sensing units. Set a uniform initial data acquisition frequency and start all smart sensor nodes to periodically collect environmental parameters; Each smart sensor node locally encapsulates the collected temperature, humidity, gas concentration, and light intensity data, adding node identifiers, collection timestamps, and location information to form raw environmental data. The raw environmental data is transmitted to the central data processing platform via a wireless sensor network.

3. The chicken farm environmental monitoring system based on intelligent sensors according to claim 2, characterized in that, The process of cleaning the raw environmental data to filter out abnormal data points caused by sensor transient failures or external transient interference, resulting in cleaned environmental data, specifically includes: The system receives the raw environmental data and parses it to obtain temperature data, humidity data, gas concentration data, light intensity data, and corresponding node identifiers, timestamps, and location information. For each environmental parameter data of each smart sensor node, the mean and standard deviation of statistical data based on a sliding time window are used; The current data point is compared with the mean of the historical data within the sliding time window. If the difference exceeds the standard deviation of a preset multiple, the current data point is determined to be an abnormal data point. Replace the values ​​of the data points marked as outliers with the mean of the historical data within the sliding time window or the values ​​generated by the interpolation algorithm. After anomaly detection and replacement of all environmental parameter data, the cleaned environmental data is output.

4. The chicken farm environmental monitoring system based on intelligent sensors according to claim 3, characterized in that, The step of time-synchronizing and aligning the cleaned environmental data, ensuring that data of the same type from different smart sensor nodes remain aligned on the time axis to form a time-aligned environmental data sequence, specifically includes: Using the system clock of the central data processing platform as a reference, calibrate the acquisition timestamp in the data packet of each smart sensor node; Set a uniform time alignment granularity to divide the timeline into continuous equal-length slices; For each time slice, iterate through all the cleaned environmental data reported by all smart sensor nodes within this time slice; If a smart sensor node has multiple data points within a certain time slice, the average value of the environmental parameter data of the smart sensor node within the time slice is taken as the representative value of the smart sensor node within this time slice. If a smart sensor node has no data points in a certain time slice, the interpolation of the data from adjacent time slices of the smart sensor node is used as the representative value of the time slice. The representative values ​​of various environmental parameters of all smart sensor nodes in the same time slice are arranged and combined to generate an environmental state snapshot of the time slice. The environmental state snapshots of all time slices are arranged in chronological order to form a time-aligned environmental data sequence.

5. The chicken farm environmental monitoring system based on intelligent sensors according to claim 4, characterized in that, The constructed environmental state feature extraction model calculates a multi-dimensional environmental feature vector representing the overall environmental state of the chicken farm based on the time-aligned environmental data sequence, specifically including: Define the dimensions of the multidimensional environmental feature vector, including spatial distribution feature dimension, temporal variation feature dimension, and parameter coupling feature dimension; Calculate spatial distribution characteristics: Select the same environmental parameter data of all smart sensor nodes in the same time slice, and calculate its global average, maximum, minimum and spatial distribution variance; Calculate the time variation characteristics: For each environmental parameter data of each smart sensor node, extract its value in multiple consecutive time slices, and calculate its short-term trend slope and fluctuation intensity. Calculate parameter coupling characteristics: Analyze the correlation coefficients between different environmental parameters on the same smart sensor node at the same time slice, such as the correlation between temperature and humidity; The spatial distribution features, temporal variation features, and parameter coupling features calculated for each time slice are concatenated to form a multidimensional environmental feature vector corresponding to the time slice.

6. The chicken farm environmental monitoring system based on intelligent sensors according to claim 5, characterized in that, The establishment of the environmental anomaly detection model, which uses the multi-dimensional environmental feature vector to calculate the environmental anomaly index at each sampling time, and determines whether the current environmental state deviates from the normal range based on the environmental anomaly index, specifically includes: Collect multidimensional environmental feature vectors from the history of chicken farms during normal production cycles to form a normal environmental feature sample set; Based on the normal environment feature sample set, a single-class classification model is trained, which can describe the distribution range of normal environment features in the feature space. For the multidimensional environmental feature vector of the current time slice obtained in real time calculation, input it into the single-class classification model to calculate its distance from the center of normal distribution or the probability of belonging to normal distribution; The distance or probability is converted into an environmental anomaly index. The higher the environmental anomaly index, the greater the degree of deviation from the normal state. A threshold for the environmental anomaly index is set. When the calculated environmental anomaly index exceeds the threshold, the current environmental state is determined to deviate from the normal range.

7. The chicken farm environmental monitoring system based on intelligent sensors according to claim 6, characterized in that, When the environmental anomaly detection model determines that the current environmental state deviates from the normal range, it triggers the environmental data analysis process, performing deep pattern mining on the environmental data during the deviation period, specifically including: Record the start and end times of environmental conditions deviating from the normal range to determine the period of deviation; Extract time-aligned environmental data sequences and corresponding multi-dimensional environmental feature vectors within the off-time period; Cluster analysis is performed on the environmental data during the off-period time period to identify a group of smart sensor nodes that exhibit similar change patterns during the off-period time period; Principal component analysis was performed on the multidimensional environmental feature vectors within the deviation period to identify the main feature dimensions that led to the increase in the environmental anomaly index. The results of the cluster analysis are correlated with the results of the principal component analysis to analyze how changes in the feature dimensions of a specific group of nodes collectively lead to anomalies in the overall environment.

8. The chicken farm environmental monitoring system based on intelligent sensors according to claim 7, characterized in that, The environmental data analysis process identifies the core environmental factors causing deviations in environmental status and their abnormal evolution patterns, and generates an environmental anomaly diagnosis report, specifically including: Based on the correlation results of the cluster analysis and principal component analysis, the environmental parameters that changed most significantly during the deviation period were identified, and these environmental parameters are the core environmental factors. Track the time series changes of the core environmental factors during the deviation period and extract their change pattern characteristics, including the time of abrupt change, the rate of change, and the duration. Analyze the interaction between the core environmental factors and other environmental parameters during the deviation period to identify abnormal coupling evolution patterns; The identification results of the core environmental factors, the description of their abnormal evolution patterns, and the analysis of their linkage with other parameters are integrated to generate a structured environmental anomaly diagnostic report.

9. A chicken farm environmental monitoring system based on intelligent sensors according to claim 8, characterized in that, The step of generating a targeted environmental control instruction sequence by matching the environmental anomaly diagnosis report with a preset environmental control strategy library specifically includes: Maintain an environmental control strategy library, which stores the mapping relationship between different environmental anomaly patterns and corresponding control measures; The core environmental factors and their abnormal evolution patterns described in the environmental anomaly diagnosis report are matched with the abnormal pattern entries in the environmental regulation strategy library. Find the abnormal pattern entries with the highest similarity and obtain the set of predefined environmental control measures associated with them; Based on the specific parameter changes in the environmental anomaly diagnosis report, the parameters in the predefined set of environmental control measures are instantiated, such as setting the target speed of the fan or the target opening of the ventilation window. The instantiated control measures are arranged according to the preset execution logic and timing to generate a specific executable environmental control instruction sequence.

10. A chicken farm environmental monitoring system based on intelligent sensors according to claim 9, characterized in that, The system also includes: The execution feedback evaluation module sends the environmental control command sequence to the environmental control execution equipment within the chicken farm, continuously receives feedback status data from the environmental control execution equipment, and evaluates and dynamically adjusts the execution effect of the environmental control command sequence in conjunction with the latest raw environmental data. Specifically, this includes: The environmental control command sequence is sent to the corresponding environmental control execution device through the control network; After executing the command, the environmental control equipment sends back its current working status as feedback status data; During the execution of the environmental control command sequence, new raw environmental data continues to be collected and processed through cleaning, alignment, feature extraction, and anomaly index calculation. Monitor the changing trends of environmental anomaly indices calculated using new raw environmental data, and evaluate the effectiveness of control measures; If the environmental anomaly index does not fall back below the normal threshold within the expected time, or if the feedback status data shows that the equipment has not executed the instructions, the environmental control strategy will be re-matched or corrected based on the latest multidimensional environmental feature vector and environmental anomaly diagnosis report, and a new environmental control instruction sequence will be generated and issued.

Images