Intelligent temperature detection system for livestock cub heat preservation house

Abstract

The invention relates to the technical field of animal house environment temperature detection, in particular to an intelligent temperature detection system for a livestock cub heat preservation house, which comprises the following steps: grouping multi-source temperature data streams collected by temperature sensors according to numbers and carrying out real-time fluctuation monitoring, so that incomplete or abnormal range deviation records can be eliminated before data enters an analysis link; the accuracy and timeliness of follow-up analysis data are guaranteed, distribution comparison is conducted on single-point temperature through an isolated forest, screening is conducted in combination with a set threshold value, abnormal points can be accurately recognized under the multi-dimensional condition, the possibility of missing detection and misjudgment is reduced, the long-short-term memory network is adopted, and the accuracy and timeliness of data analysis are improved. According to the method, the time sequence modeling of the long-short term dependency relationship is introduced to calculate the temperature change rate, and the future short-term temperature prediction is generated when the continuous change rate reaches the standard, so that the possible temperature anomaly trend can be found in advance, and the rapid and accurate temperature anomaly management capability in a complex environment is realized; and the response speed and the processing accuracy of the sudden temperature control problem are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of livestock environment temperature detection, and in particular to a livestock young animal incubator temperature intelligent detection system. BACKGROUND

[0002] The livestock environment temperature detection technical field is applied to the environmental monitoring equipment of livestock farms and breeding workshops, is used for real-time collection, analysis and feedback of livestock internal air temperature data, relates to livestock environment regulation and control, animal physiological optimum temperature zone management and sensor-based data monitoring, and aims to guarantee that young livestock at different growth stages are bred under suitable temperature conditions, and avoid slow growth or increased mortality rate caused by environmental temperature fluctuations.

[0003] A livestock young animal incubator temperature intelligent detection system aims to realize real-time and accurate monitoring and analysis of the internal temperature of a livestock young animal incubator, and automatically drive temperature control equipment according to the detection result, so that the temperature in the incubator is continuously maintained in the optimal physiological temperature range of young animals, thereby reducing the stress reaction and disease risk caused by abnormal temperature, improving the survival rate and growth rate of young animals, and guaranteeing the production benefit of breeding.

[0004] In the prior art, temperature monitoring usually relies on sensor timed collection and basic threshold judgment, lacks real-time screening of collected data quality, and error data caused by sensor failure and environmental interference is easy to enter the analysis link, causing misjudgment or delayed response. Abnormal identification relies on single upper and lower limit judgment, lacks in-depth analysis of dynamic characteristics and multi-mode abnormalities of temperature changes, cannot accurately distinguish different risk types such as rapid change and long-time deviation, and the prior art cannot accurately map data abnormalities and spatial positions, resulting in low problem tracing efficiency, trend judgment generally stays in the static data comparison stage, lacks prediction ability for future temperature changes, and is easy to miss the opportunity for early intervention. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art and provide a livestock young animal incubator temperature intelligent detection system.

[0006] In order to achieve the above purpose, the present application adopts the following technical scheme: a livestock young animal incubator temperature intelligent detection system comprises: A temperature data real-time monitoring module: acquires real-time temperature data streams of each sensor in the livestock incubator, divides data sets according to sensor numbers, monitors temperature fluctuations in real time, removes invalid data, retains valid data points, saves valid data records, and generates real-time temperature change records; A temperature anomaly detection module: based on the real-time temperature change records, compares each data point with a preset temperature range by using an isolation forest, marks as abnormal if it is out of range, uses a set threshold to filter abnormal data, and generates an abnormal data point set; Anomaly pattern recognition module: based on the set of abnormal data points, the degree of abnormality of each data point is determined, and if it exceeds the set score threshold, it is classified as "local rapid warming", "local rapid cooling" and "long-term deviation from the temperature zone", and the abnormal pattern classification result is generated; Abnormal cause analysis module: based on the abnormal pattern classification result, the geographic location of each abnormal data point is obtained, the abnormal position is identified by matching the coordinates with the livestock shed structure map, the possible causes are analyzed according to the cause mapping table, and the cause speculation report is generated; Temperature trend prediction and evolution module: based on the real-time temperature change record, a long short-term memory network is used to calculate the temperature change rate in a fixed time window, if the temperature change rate is positive and exceeds the set value in the last three windows, it is determined that the temperature is rising, the temperature change in the next 5 to 10 minutes is predicted, and the temperature change evolution trend is generated.

[0007] As a further scheme of the application, the temperature data real-time monitoring module comprises: Temperature data acquisition submodule: based on the temperature data stream of each sensor in the livestock shed, real-time data acquisition is performed according to the sensor number, the temperature data of each sensor is obtained, and each data point is marked with a time stamp, real-time transmission of the temperature data stream is performed, continuous updating of each sensor data is ensured, and a real-time temperature data stream is generated; Data validity screening submodule: based on the real-time temperature data stream, the data of each sensor is judged for validity, and missing or unreasonable temperature data such as data points exceeding the maximum and minimum range, incomplete acquisition, etc. are removed, and valid temperature data points are retained, and all valid data are classified to generate a valid temperature data set.

[0008] As a further scheme of the application, the temperature anomaly detection module comprises: Abnormal data marking submodule: based on the real-time temperature change record, the temperature data is input into the isolated forest model, each data point is judged for abnormality, the distribution rule of normal data is learned according to historical data in the training stage, the abnormal score of each data point is calculated, and if the data point exceeds the set maximum and minimum value, it is immediately marked as abnormal, and an abnormal data marking result is generated; Abnormal data screening submodule: based on the abnormal data marking result, all data points marked as abnormal are screened, the validity of the abnormal data is further judged by setting a threshold, invalid, repeated or non-standard abnormal points in the data are removed, and a screened abnormal data set is generated; Abnormal data point set generation submodule: based on the screened abnormal data set, all valid abnormal data points are sorted, classified, and summarized into a complete data set according to their characteristics, and an abnormal data point set is generated.

[0009] As a further scheme of the present application, the Isolation Forest, firstly, inputs the temperature data obtained based on the real-time temperature change record in the order of sensor number and time sequence into the Isolation Forest model, constructs multiple Isolation Trees randomly partitioning the feature space in the training stage by using the historical temperature data, splits the data points to the leaf nodes by recursively randomly selecting the temperature features and randomly selecting the split points in the feature value range, records the path length of each data point in each Isolation Tree, calculates the average Isolation Depth, then calculates the theoretical expected Isolation Depth according to the size of the data set, normalizes the average Isolation Depth to obtain the anomaly score of each data point, and immediately marks the data point as abnormal when the anomaly score of the data point exceeds the set threshold or the temperature value of the data point directly exceeds the preset temperature peak and valley range, and outputs the abnormal data marking result.

[0010] As a further scheme of the present application, the Isolation Forest, according to the formula:

[0011] Among them: represents the anomaly score of the data point , represents the average Isolation Depth required by the data point in the Isolation Forest model, represents the size of the data set , represents the weight coefficient of the data point , represents the density coefficient of the data point , represents the change coefficient of the data point , controls the influence of on the anomaly score, represents the density influence coefficient, controls the influence of on the anomaly score, represents the change coefficient adjustment coefficient.

[0012] As a further scheme of the present application, the anomaly pattern recognition module comprises: an anomaly classification submodule: based on the set of abnormal data points, the anomaly score of each data point is determined, and by comparing the anomaly value of the data point with the preset threshold, if it exceeds the set standard, it is classified, and the abnormal data points are classified into three categories of “local rapid temperature rise”, “local rapid temperature drop” and “long-time deviation from temperature zone”, and a classification result is generated; a pattern recognition result generation submodule: based on the classification result, the abnormal data points are summarized according to the type, the integration of the classified data is performed, the abnormal pattern is sorted and classified, and an abnormal pattern classification result is generated.

[0013] As a further aspect of the present invention, the anomaly cause analysis module includes: Location identification submodule: Based on the abnormal pattern classification results, obtain the geographic coordinates of each abnormal data point, match the sensor location with the livestock shed structure map, identify the specific location of each abnormal data point, and generate an abnormal location dataset.

[0014] Cause prediction submodule: Based on the abnormal location dataset, analyze each abnormal pattern, compare the abnormal data points with potential causes by consulting the cause mapping table, analyze possible causes such as heating equipment overload and temperature leakage, and generate cause prediction results.

[0015] As a further aspect of the present invention, the temperature trend prediction and evolution module includes: Temperature change rate calculation submodule: Based on the real-time temperature change records, multidimensional climate data is organized into time series data, input into a long short-term memory network model for training and prediction, and combined with the temporal correlation of historical data, the temperature change rate within each time window is calculated, and the change rate for each time period is output to generate temperature change rate results. Temperature rise trend judgment submodule: Based on the temperature change rate result, compare the temperature change rate of three consecutive time windows. If the temperature change rate in each window is positive and exceeds the set threshold, it is judged as a temperature rise trend, and a temperature rise trend judgment result is generated. Temperature change prediction submodule: Based on the temperature rise trend determination result, predict the temperature change in the next 5 to 10 minutes, calculate the temperature value for the future time period according to the temperature rise trend, and generate temperature change prediction result.

[0016] As a further aspect of the present invention, firstly, the multidimensional climate time series data based on real-time temperature change records is divided into fixed time windows and normalized to form an input sequence vector. At each time step of the network, the input sequence, along with the hidden state and unit state of the previous time step, enters the forget gate, input gate, and output gate. The forget gate outputs the forget coefficient based on the calculation results of the weight matrix and bias term. The input gate generates the input coefficient and updates the unit state in combination with the candidate state vector. The output gate calculates the hidden state based on the updated unit state and uses it as the input for the next time step. The network captures the time dependence of temperature changes through multi-time-step iterative calculations. In the backpropagation stage, the weight coefficients and bias terms of each gate are updated using the time backpropagation algorithm to minimize the prediction error. The updated network weights are used to perform forward calculations on future time windows, outputting the temperature change rate within each time window. Based on these rate results, the warming trend is determined and the temperature is predicted for the next 5 to 10 minutes.

[0017] As a further scheme of the present application, the long short-term memory network is according to the formula:

[0018] Wherein: represents the temperature change rate at time t, represents the temperature at time t, represents the temperature at time t, represents the time interval of the time window, represents the weight coefficient, represents the density coefficient, represents the adjustment factor, represents the climate change coefficient.

[0019] Compared with the prior art, the present application has the advantages and positive effects that: In the present application, by grouping the multi-source temperature data stream collected by the temperature sensor according to the number and performing real-time fluctuation monitoring, incomplete or abnormally deviated records can be eliminated before the data enters the analysis link, ensuring the accuracy and timeliness of the subsequent analysis data; In the present application, the isolated forest is used to compare the distribution of single-point temperature and combined with the set threshold for screening, which can accurately identify abnormal points under multi-dimensional conditions and reduce the possibility of missed detection and misjudgment. The abnormal data screened out are divided into various specific modes through abnormal degree determination, which is helpful for targeted analysis of the type and range of problem occurrence. In the abnormal position analysis, the spatial positioning is realized by combining the abnormal data coordinates with the layout of the livestock house structure, and the reason inference is realized through the cause mapping table, which can directly associate the environment and the running state of the equipment; In the present application, the long short-term memory network is used to introduce the long short-term dependence relationship of the time series modeling calculation of the temperature change rate, and the future short-time temperature prediction is generated when the continuous change rate meets the standard, which can find the possible temperature abnormal trend in advance, realize the fast, accurate and traceable temperature abnormal management ability in complex environment, and improve the response speed and processing accuracy in dealing with sudden temperature control problems. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 The system flowchart of the present application. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application is further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application, and are not used to limit the present application.

[0022] In the description of the present application, it should be understood that the terms "length", "width", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, in the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly and specifically limited.

[0023] Embodiment one Please refer to Figure 1 The present application provides a technical solution: an intelligent temperature detection system for livestock young animal incubator, comprising: A temperature data real-time monitoring module: acquires real-time temperature data streams of each sensor in the livestock house, divides the data set according to the sensor number, monitors the temperature fluctuation in real time, eliminates invalid data, retains valid data points, saves valid data records, and generates real-time temperature change records; A temperature anomaly detection module: based on the real-time temperature change records, compares each data point with the preset temperature interval by using the Isolation Forest, and if it exceeds the range, it is marked as abnormal, and the abnormal data is screened using a set threshold, and an abnormal data point set is generated; An abnormal pattern recognition module: based on the abnormal data point set, the abnormal degree of each data point is determined, and if it exceeds the set score threshold, it is classified as "local rapid heating", "local rapid cooling" and "long time deviation from temperature zone", and an abnormal pattern classification result is generated; An abnormal cause analysis module: based on the abnormal pattern classification result, the geographic position of each abnormal data point is obtained, the abnormal position is identified by matching the coordinates with the livestock house structure diagram, the possible causes are analyzed according to the cause mapping table, and a cause speculation report is generated; A temperature trend prediction and evolution module: based on the real-time temperature change records, a long short-term memory network is used to calculate the temperature change rate in a fixed time window, and if the temperature change rate is positive and exceeds the set value in the last three windows, it is determined as a heating trend, the temperature change in the next 5 to 10 minutes is predicted, and a temperature change evolution trend is generated.

[0024] The temperature data real-time monitoring module comprises: A temperature data acquisition sub-module: based on the temperature data stream of each sensor in the livestock house, real-time data acquisition is performed according to the sensor number, the temperature data of each sensor is acquired, and each data point is marked according to the time stamp, real-time transmission of the temperature data stream is performed, continuous updating of each sensor data is ensured, and a real-time temperature data stream is generated; Data validity screening submodule: based on real-time temperature data stream, the data of each sensor is judged for validity, and missing or unreasonable temperature data is removed, such as data points exceeding the maximum and minimum range, incomplete collection, etc. Effective temperature data points are retained, and all effective data are classified to generate an effective temperature data set. Temperature data acquisition submodule: based on the temperature data stream output by each sensor in the livestock house, the TCP / IP socket long connection transmission method is used to create a socket at the acquisition end, with parameters including address family AF_INET, socket type SOCK_STREAM, protocol 0. Data channels are established in order for sensor numbers, and the port number 5020 and IP address 192.168.1.100 are bound at the receiving end. The maximum number of listening connections is set to 50, connection requests are received and file descriptors are generated, and a fixed-size buffer of 1024 bytes is used to receive data packets in a loop. Each received data point is timestamped with a value composed of seconds and milliseconds, and the sensor number and timestamp are bound into a key-value pair and stored in a two-dimensional array structure. The marked data is sent to the server cache queue in the form of a data packet. The cache queue uses a ring buffer structure with a capacity of ten thousand data units. The data unit structure includes sensor number, timestamp, and temperature value. All data is written to the buffer in a first-in, first-out order using stream transmission, generating a real-time temperature data stream. Data validity screening submodule: based on real-time temperature data stream, the Z-score statistical method is used for validity judgment. First, calculate the arithmetic mean of the temperature values of each sensor in the last sixty seconds, and use the average value as the reference temperature. Then calculate the average of the square of the difference between the temperature value and the reference temperature in the time period and take the square root to get the standard deviation. Divide the difference between the temperature value of a single data point and the reference temperature by the standard deviation to get the standardized value. If the absolute value of the standardized value is greater than three, the data point is removed from the data queue. Missing values are judged and removed if they are missing. Temperature data points higher than 45 or lower than 0 are filtered out by conditional judgment. When removing incomplete data, check the number of fields in a single data, and if it is less than three, it is directly removed. All retained data points are classified by sensor number and packaged as key-value pairs in JSON format, with the key being the sensor number and the value being a list of temperature records sorted in ascending order of timestamp, generating an effective temperature data set.

[0025] The temperature anomaly detection module includes: Abnormal data marking submodule: based on real-time temperature change records, the temperature data is input into the Isolation Forest model, and each data point is judged for abnormality. In the training stage, the distribution rule of normal data is learned from historical data, and the abnormal score of each data point is calculated. If the data point exceeds the set maximum and minimum value, it is immediately marked as abnormal, and an abnormal data marking result is generated. Anomaly data screening submodule: based on the anomaly data marking result, all data points marked as abnormal are screened, the effectiveness of the abnormal data is further judged by setting a threshold, invalid, repeated or non-standard abnormal points in the data are removed, and a screened abnormal data set is generated; Anomaly data point set generation submodule: based on the screened abnormal data set, all valid anomaly data points are sorted, classified, and summarized into a complete data set according to their characteristics, and an anomaly data point set is generated; Anomaly data marking submodule: based on real-time temperature change records, an isolation forest model is used to determine the anomaly of each data point, historical temperature data is used as a training set to input the model training stage, the number of isolation trees is set to 200, the sample sampling ratio is 0.5, and the feature sampling ratio is 1.0. In the process of building each isolation tree, a temperature feature field is randomly selected and a split threshold is randomly determined, and the data is recursively divided until the sample size is 1 or the maximum depth is reached. The path length of each data point in all isolation trees is recorded and the average path depth is calculated. The average path depth is compared with the expected path depth of the same scale data set to generate an anomaly score. If the data point temperature value is greater than the set maximum value 45 or less than the set minimum value 0, the data point is directly marked as abnormal, and an anomaly data marking result is generated. Anomaly data screening submodule: based on the anomaly data marking result, a threshold screening method is used to judge the effectiveness of the data points marked as abnormal. The anomaly score threshold is set to 0.65. All anomaly data points are compared with the threshold value. The data points with a score lower than the threshold value are removed. The data deduplication program is called to detect and delete duplicate records using temperature value, sensor number, and timestamp as joint keys. The data point field integrity is detected and records missing temperature value, sensor number, or timestamp are deleted. The remaining data points that meet the standard are stored in a new data set, and a screened abnormal data set is generated. Anomaly data point set generation submodule: based on the screened abnormal data set, a classification sorting method is used to classify all valid anomaly data points by temperature change amplitude interval, change duration, and occurrence location. An independent index number is generated for each classification and a mapping table is established. The same type of anomaly data points are grouped into the same index group. The grouped data is stored in a structured key-value pair form, with the key as the index number and the value as the complete record set of the corresponding anomaly data point. An anomaly data point set is generated.

[0026] The isolated forest, firstly, the temperature data obtained based on the real-time temperature change record is input into the isolated forest model according to the sensor number and time sequence, and the historical temperature data is used to construct multiple isolated trees with randomly divided feature spaces in the training stage, the data points are split to the leaf nodes by recursively randomly selecting the temperature features and randomly selecting the split points in the feature value range, the path length of each data point in each isolated tree is recorded, the average isolated depth is calculated, then the theoretical expected isolated depth is calculated according to the size of the data set, the average isolated depth is normalized to obtain the anomaly score of each data point, when the anomaly score of the data point exceeds the set threshold or the temperature value directly exceeds the preset temperature peak and valley range, the data point is immediately marked as abnormal, and the abnormal data marking result is output; The isolated forest, according to the formula:

[0027] Among them: The anomaly score of the data point , The average isolated depth required by the data point in the isolated forest model, When the size of the data set is , The weight coefficient of the data point , The density coefficient of the data point , The change coefficient of the data point , control The influence of the anomaly score, The density influence coefficient, control The influence of the anomaly score, The change coefficient adjustment coefficient; Execution process: firstly, the isolated forest model calculates the average isolated depth of each data point , which indicates the difficulty of isolating the data point in the isolated forest, and the point with shallow isolated depth indicates higher abnormal degree, then the isolated forest model calculates the expected depth according to the size of the data set , which is used to normalize the anomaly score of the data point, then the anomaly score of the data point is calculated, wherein It is an exponential decay function based on the isolated depth and the expected depth , used to measure the abnormal degree of the data point, and the weight coefficient By adjusting the importance of the data points, the influence of the data points on the final anomaly score is reflected, and the points with greater weights will have a more significant impact on the anomaly score, the density coefficient By calculating the density of the environment around the data points, the possibility of anomaly is reflected, and the area with higher density is usually more likely to produce abnormal data points, the change coefficient The change speed of the data points in the time dimension is considered, and the data points with faster changes may represent the rapid fluctuations of the temperature, and the coefficient in the final formula , β, γ are preset weight adjustment parameters for controlling the influence degree of the new factors on the anomaly score, and the specific adjustment coefficient value can be set according to the characteristics of the actual temperature control data. Through the introduction of the parameters, the improved formula can more accurately detect the abnormal fluctuations of the temperature in the heat preservation shed, and ensure that the growth environment of the cubs is always maintained in the suitable temperature range.

[0028] The anomaly pattern recognition module comprises: The anomaly classification submodule: based on the set of abnormal data points, the anomaly score of each data point is determined, and by comparing the abnormal value of the data point with the preset threshold, if it exceeds the set standard, it is classified, and the abnormal data points are classified into three categories of “local rapid heating”, “local rapid cooling” and “long time deviating from the temperature zone”, and the classification result is generated. The pattern recognition result generation submodule: based on the classification result, the abnormal data points are summarized according to the type, the integration of the classified data is carried out, the anomaly pattern is sorted and classified, and the anomaly pattern classification result is generated. The anomaly classification submodule: based on the set of abnormal data points, a multi-class logistic regression classification method is used to determine the anomaly score of each data point, the input data is formatted into a three-dimensional feature vector containing the anomaly score, the temperature change amplitude and the change duration, the number of classification categories is set to 3, corresponding to the three categories of local rapid heating, local rapid cooling and long time deviating from the temperature zone, the weight matrix and the bias vector are fitted in the training stage using the historical labeled data, the learning rate is set to 0.01, the number of iterations is set to 500, the prediction probability is calculated and the weight coefficient is updated by batch gradient descent in each iteration, and the cross-entropy loss function is used to evaluate the classification error. In the classification stage, the three-dimensional feature vector of each data point is input into the trained classification model, the probability values of the three categories are calculated, the category with the largest probability value is selected as the classification label of the data point, and the classification result is generated. The mode recognition result generation submodule: based on the classification result, the abnormal data points are aggregated by category type using a grouping aggregation statistical method. An index table containing three types of category labels is first established. The abnormal data points of the same category are arranged in ascending order of timestamp and stored in the corresponding category group. The count function is called to count the number of abnormal data points in the category. The time span calculation function is called to obtain the time interval of the first and last data points in the category. The amplitude range calculation function is used to determine the maximum and minimum values of the temperature change of the abnormal data points in the category. The above statistical information and the corresponding abnormal data records are stored as structured data units. Finally, the structured data units of all groups are combined in order of category to generate the abnormal mode classification result.

[0029] The abnormal cause analysis module includes: The location recognition submodule: based on the abnormal mode classification result, the geographic coordinates of each abnormal data point are obtained. The sensor location is matched with the livestock house structure diagram to identify the specific location of each abnormal data point and generate an abnormal location data set.

[0030] The cause speculation submodule: based on the abnormal location data set, each abnormal mode is analyzed by consulting the cause mapping table to compare the abnormal data points with potential causes, analyze possible causes such as overheating of heating equipment, temperature leakage, etc., and generate a cause speculation result. The location recognition submodule: based on the abnormal mode classification result, the geographic coordinates of each abnormal data point are obtained. The sensor location is matched with the livestock house structure diagram to identify the specific location of each abnormal data point and generate an abnormal location data set. The cause speculation submodule: based on the abnormal location data set, each abnormal mode is analyzed by consulting the cause mapping table to compare the abnormal data points with potential causes, analyze possible causes such as overheating of heating equipment, temperature leakage, etc., and generate a cause speculation result.

[0031] The temperature trend prediction and evolution module comprises: The temperature change rate calculation submodule: based on the real-time temperature change record, the multi-dimensional climate data is organized into time series data, input into the long short-term memory network model for training and prediction, the time correlation of historical data is combined, the temperature change rate in each time window is calculated, and the change rate of each time period is output to generate the temperature change rate result; The temperature change rate calculation submodule: based on the real-time temperature change record, the multi-dimensional climate data is organized into time series data, input into the long short-term memory network model for training and prediction, the time correlation of historical data is combined, the temperature change rate in each time window is calculated, and the change rate of each time period is output to generate the temperature change rate result; The temperature change prediction submodule: based on the temperature trend determination result, the temperature change in the next 5 to 10 minutes is predicted, the temperature value of the future time period is calculated according to the temperature trend, and the temperature change prediction result is generated; The temperature change rate calculation submodule: based on the real-time temperature change record, the multi-dimensional climate data is organized into time series data, input into the long short-term memory network model for training and prediction, the time correlation of historical data is combined, the temperature change rate in each time window is calculated, and the change rate of each time period is output to generate the temperature change rate result; The temperature change rate calculation submodule: based on the real-time temperature change record, the multi-dimensional climate data is organized into time series data, input into the long short-term memory network model for training and prediction, the time correlation of historical data is combined, the temperature change rate in each time window is calculated, and the change rate of each time period is output to generate the temperature change rate result; Temperature Change Prediction Submodule: Based on the temperature rise trend determination result, the Long Short-Term Memory (LSTM) network prediction method is used to calculate the future temperature. The temperature value sequence of continuous time windows is used as the input vector, with a vector length of twelve steps, corresponding to a time span of one minute. The vector is input into the pre-trained LSM network model for forward propagation, predicting the temperature value for each minute within the next five to ten minutes. During the prediction process, the hidden state and the unit state are continuously transmitted. The number of output layer nodes is the number corresponding to the prediction time step of five to ten. The output temperature values ​​are arranged in chronological order and retained to two decimal places in floating-point form, generating the temperature change prediction result.

[0032] First, the multidimensional climate time series data, compiled based on real-time temperature change records, is divided into fixed time windows and normalized to form an input sequence vector. At each time step of the network, the input sequence, along with the hidden state and unit state of the previous time step, enters the forget gate, input gate, and output gate. The forget gate outputs the forget coefficient based on the calculation results of the weight matrix and bias term. The input gate generates input coefficients and updates the unit state in combination with the candidate state vector. The output gate calculates the hidden state based on the updated unit state and uses it as the input for the next time step. The network captures the time dependence of temperature changes through multi-time-step iterative calculations. In the backpropagation stage, the weight coefficients and bias terms of each gate are updated using the time backpropagation algorithm to minimize the prediction error. The updated network weights are used to perform forward calculations on future time windows, outputting the temperature change rate within each time window. Based on these rate results, the warming trend is determined and the temperature is predicted for the next 5 to 10 minutes. Long Short-Term Memory (LSTM) networks, according to the formula:

[0033] in: Indicates time The rate of temperature change at any given time Indicates time Temperature at any moment Indicates time Temperature at any moment This represents the time interval of the time window. Indicates the weighting coefficient. Represents the density coefficient. Indicates the regulating factor. Indicates the climate change coefficient; Execution process: First, calculate the time. and time Temperature difference over time This reflects the magnitude of temperature change within a time window, and the magnitude of temperature change is related to the weighting coefficient. Multiplication, weighting coefficients For adjusting its influence on the change rate according to different temperature intervals, if the temperature fluctuation is within the critical range, the weight coefficient will increase to enhance the sensitivity of the change rate, then the temperature difference and the density coefficient are multiplied, the density coefficient is calculated according to the temperature data distribution in the current time period, in the temperature interval with high density, the influence of temperature change is larger, the change rate of this time period is improved, then the time interval is used as a normalization factor to ensure the consistency of temperature change rate and time span, avoid the calculation deviation caused by different time window sizes, finally introduce external climate factors, adjust the factor and the climate change coefficient are used to adjust the influence of climate factors on the temperature change rate, the adjustment factor controls the overall influence intensity of the climate factor, the climate change coefficient is dynamically adjusted according to the influence of real-time climate data on temperature change, which enhances the adaptability of the model to climate change, the formula comprehensively considers multiple factors such as temperature change, external climate factors, data distribution, etc., so as to accurately calculate the temperature change rate in each time window, help to monitor and adjust the temperature of the temperature preservation house in real time, and ensure that the cubs are in a suitable temperature environment.

[0034] The above is only the preferred embodiment of the present application, not other forms of the present application, any skilled in the art may use the above disclosed technical content to change or modify equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification of the above embodiments without departing from the technical solution content of the present application, according to the technical essence of the present application, still belongs to the protection scope of the technical solution of the present application.

Claims

1. A livestock young animal incubator temperature intelligent detection system, characterized in that, The system comprises: A temperature data real-time monitoring module: acquires real-time temperature data streams of each sensor in the livestock house, divides the data set according to the sensor number, monitors the temperature fluctuation in real time, eliminates invalid data, retains valid data points, saves valid data records, and generates real-time temperature change records; A temperature anomaly detection module: based on the real-time temperature change records, compares each data point with the preset temperature interval through the isolation forest, and if it exceeds the range, it is marked as abnormal; uses a set threshold to filter abnormal data, and generates an abnormal data point set; An abnormal pattern recognition module: based on the abnormal data point set, determines the abnormality degree of each data point, and if it exceeds the set score threshold, it is classified as "local rapid heating", "local rapid cooling", "long-time deviation from the temperature zone", and an abnormal pattern classification result is generated; An abnormal cause analysis module: based on the abnormal pattern classification result, acquires the geographic location of each abnormal data point, matches the coordinates with the livestock house structure diagram, identifies the abnormal position, analyzes the possible causes according to the cause mapping table, and generates a cause speculation report; A temperature trend prediction and evolution module: based on the real-time temperature change records, adopts a long short-term memory network, calculates the temperature change rate in a fixed time window, if the temperature change rate in the last three windows is positive and exceeds the set value, it is determined as a heating trend, the temperature change in the next 5 to 10 minutes is predicted, and a temperature change evolution trend is generated.

2. The livestock young animal temperature detection system according to claim 1, characterized in that, The temperature data real-time monitoring module comprises: A temperature data acquisition sub-module: based on the temperature data stream of each sensor in the livestock house, real-time data acquisition is performed according to the sensor number, the temperature data of each sensor is acquired, and each data point is marked with a timestamp, real-time transmission of the temperature data stream is performed, the continuous update of each sensor data is ensured, and a real-time temperature data stream is generated; A data validity filtering sub-module: based on the real-time temperature data stream, the validity of the data of each sensor is judged, invalid or unreasonable temperature data such as data points exceeding the maximum and minimum range, incomplete acquisition, etc. is eliminated, valid temperature data points are retained, and all valid data are classified to generate a valid temperature data set.

3. The livestock young animal temperature detection system of claim 1, wherein The temperature anomaly detection module comprises: An abnormal data marking sub-module: based on the real-time temperature change records, the temperature data is input into the isolation forest model, each data point is judged for abnormality, the distribution rule of normal data is learned according to historical data in the training stage, the abnormal score of each data point is calculated, and if the data point exceeds the set maximum and minimum value, it is immediately marked as abnormal, and an abnormal data marking result is generated; An abnormal data filtering sub-module: based on the abnormal data marking result, all data points marked as abnormal are filtered, the validity of the abnormal data is further judged through a set threshold, invalid, repeated or non-standard abnormal points in the data are eliminated, and a filtered abnormal data set is generated; An abnormal data point set generation sub-module: based on the filtered abnormal data set, all valid abnormal data points are sorted, classified, and summarized into a complete data set according to their characteristics, and an abnormal data point set is generated.

4. The livestock young animal temperature detection system of claim 3, wherein, The isolated forest firstly inputs the temperature data obtained based on the real-time temperature change record according to the sensor number and time sequence into the isolated forest model, constructs multiple isolated trees with randomly divided feature spaces in the training stage by using historical temperature data, splits the data points to leaf nodes by recursively randomly selecting temperature features and randomly selecting split points in the feature value range, records the path length of each data point in each isolated tree, calculates the average isolated depth, then calculates the theoretical expected isolated depth according to the size of the data set, normalizes the average isolated depth to obtain the anomaly score of each data point, and immediately marks the data point as abnormal when the anomaly score of the data point exceeds the set threshold or the temperature value of the data point directly exceeds the preset temperature peak and valley range, and outputs the abnormal data marking result.

5. The livestock young animal temperature detection system according to claim 3, characterized in that, The isolated forest is according to the formula: in: Representing data points Abnormal scores, Representing data points The required average isolation depth in the isolation forest model Indicates the dataset size is hour, Representing data points The weighting coefficients, Representing data points density coefficient, Representing data points The coefficient of change, control Impact on outlier scores Represents the density influence coefficient, controlling Impact on outlier scores This represents the adjustment coefficient for the change.

6. The livestock young animal temperature detection system of claim 1, wherein, The abnormal pattern recognition module comprises: An abnormal classification submodule: based on the abnormal data point set, the anomaly score of each data point is determined, and the anomaly value of the data point is compared with the preset threshold value, if it exceeds the set standard, classification is performed, the abnormal data points are classified into "local rapid heating", "local rapid cooling" and "long-time deviation from temperature zone", and a classification result is generated; A pattern recognition result generation submodule: based on the classification result, the abnormal data points are summarized according to types, the integration of classified data is performed, the abnormal pattern is sorted and classified, and an abnormal pattern classification result is generated.

7. The livestock young animal temperature detection system of claim 1, wherein, The abnormal cause analysis module comprises: A position recognition submodule: based on the abnormal pattern classification result, the geographic coordinates of each abnormal data point are obtained, the sensor position is matched with the livestock house structure diagram, the specific position of each abnormal data point is identified, and an abnormal position data set is generated. A cause speculation submodule: based on the abnormal position data set, each abnormal pattern is analyzed, the abnormal data points are compared with potential causes by consulting a cause mapping table, possible causes such as overheating of heating equipment and temperature leakage are analyzed, and a cause speculation result is generated.

8. The livestock young animal temperature detection system of claim 1, wherein, The temperature trend prediction and evolution module comprises: A temperature change rate calculation submodule: based on the real-time temperature change record, multi-dimensional climate data is organized into time series data, input into a long short-term memory network model for training and prediction, the time correlation of historical data is combined, the temperature change rate in each time window is calculated, and the change rate of each time period is output, and a temperature change rate result is generated; A heating trend judgment submodule: based on the temperature change rate result, the temperature change rates of three consecutive time windows are compared, if the temperature change rate in each window is positive and exceeds the set threshold, it is determined as a heating trend, and a heating trend judgment result is generated; A temperature change prediction submodule: based on the heating trend judgment result, the temperature change in the next 5 to 10 minutes is predicted, the temperature value of the future time period is calculated according to the heating trend, and a temperature change prediction result is generated. 9.The livestock youngling incubator temperature intelligent detection system according to claim 8, characterized in that, The long short-term memory network first divides the multi-dimensional climate time series data based on the real-time temperature change record into fixed time windows and performs normalization processing to form an input sequence vector. At each time step of the network, the input sequence, the hidden state of the previous time step and the cell state are jointly input into the forget gate, the input gate and the output gate. The forget gate outputs a forgetting coefficient according to the operation result of the weight matrix and the bias term. The input gate generates an input coefficient and updates the cell state in combination with the candidate state vector. The output gate calculates the hidden state according to the updated cell state and uses the hidden state as the input of the next time step. The network captures the time dependence of the temperature change through the multi-time step cycle calculation. In the backward propagation stage, the time backward propagation algorithm is used to update the weight coefficients and the bias terms of each gate to minimize the prediction error. The updated network weight is used to perform forward calculation on the future time window, and the temperature change rate in each time window is output. Based on the rate results, the temperature rising trend is determined and the temperature in the next 5 to 10 minutes is predicted.

10. The livestock youngling incubator temperature intelligent detection system according to claim 8, characterized in that, The long short-term memory network is according to the formula: wherein: represents the time temperature rate of change at the time represents the time temperature at the time represents the time temperature at the time represents the time interval of the time window represents the weight coefficient represents the density coefficient represents the adjustment factor represents the climate change coefficient.

Images