Grain storage management system based on Internet of Things technology

Abstract

The invention relates to the technical field of supply chain management, in particular to a grain storage management system based on the Internet of Things technology, which comprises an environmental parameter monitoring module, a parameter screening module, a dynamic modeling module, an anomaly detection module and an optimal configuration module. According to the method, key environment parameters are screened through joint probability distribution and mutual information calculation, factors influencing the grain storage quality are accurately recognized, risk assessment pertinence is enhanced, short-time dynamic adjustment and trend fitting are carried out on different levels of parameters through layered dynamic modeling, it is ensured that prediction can quickly respond to short-term changes and can also reflect long-term trends, and the prediction accuracy is improved. Abnormality detection is combined with residual analysis and dynamic threshold adjustment, the accuracy of abnormality early warning is improved, the calculation frequency is adjusted according to storage stage characteristics in the optimal configuration link, parameter screening weights are redistributed based on the environment state, the regulation and control precision is improved, energy consumption is reduced, environment monitoring, abnormality detection and configuration strategies are optimized through the overall scheme, and grain storage is safer. And the loss is lower.

Description

Technical Field

[0001] The present invention relates to the technical field of supply chain management, and in particular to a grain storage management system based on Internet of Things technology. Background Art

[0002] The field of supply chain management technology includes planning, execution, monitoring and optimization of all links in the supply chain, covering the entire process from raw material procurement, inventory management, warehousing, transportation, distribution to end-user delivery. The core content of this technology field involves the coordinated management of information flow, logistics and capital flow, aiming to improve resource utilization, reduce operating costs and increase response speed. Through computer systems, the Internet of Things, big data analysis and other technical means, this field realizes data collection, analysis and intelligent decision-making in all links of the supply chain, promoting the development of intelligence, visualization and automation.

[0003] Among them, the grain storage management system based on the Internet of Things technology refers to the use of the Internet of Things technology to conduct real-time monitoring and intelligent management of environmental monitoring, storage scheduling, quality traceability, inventory management and logistics coordination in the grain storage process. The system uses a wireless sensor network to obtain key parameters such as temperature, humidity, and gas composition of the storage environment, and records the storage information of grain through radio frequency identification technology. At the same time, it combines the cloud computing platform for data processing to achieve real-time analysis and remote management of storage status. In addition, the system adjusts key links such as ventilation systems and temperature control equipment through automated control equipment to ensure the safety of grain storage, and optimizes the grain in and out of storage processes through intelligent scheduling to improve the accuracy and efficiency of storage management.

[0004] In the processing of environmental monitoring data, existing technologies rely on fixed parameter screening for risk identification, which makes it difficult to adjust monitoring priorities according to environmental changes, and reduces adaptability to emergencies. The modeling method is mainly static, which fails to effectively capture short-term environmental fluctuations and long-term trend changes, resulting in large prediction errors. Anomaly detection relies on fixed thresholds, which are difficult to adjust in real time according to environmental conditions, affecting the accuracy of anomaly warnings. The optimization configuration fails to adjust the calculation strategy in accordance with the characteristics of different storage stages, and the parameter control method is fixed, which affects the accuracy of management decisions and increases energy consumption and management costs. These deficiencies may make it difficult to accurately control the grain storage environment in long-term operation, increase the risk of mildew, and affect grain quality and storage efficiency. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a grain storage management system based on the Internet of Things technology.

[0006] In order to achieve the above purpose, the present invention adopts the following technical solution: A grain storage management system based on Internet of Things technology includes: The environmental parameter monitoring module collects grain pile temperature, grain surface humidity gradient, and gas concentration parameters in the storage environment in real time through sensors, performs sliding average calculation and standardization on the data, and generates multi-parameter standardized observation values; The parameter screening module calls the multi-parameter standardized observation value, calculates the joint probability distribution of the grain pile temperature and the grain surface humidity gradient, calculates the mutual information value of the gas concentration parameter and the target variable, and screens the parameters with a correlation with the risk of mildew in the grain pile higher than the threshold by comparing the mutual information value with the set threshold, thereby generating a core parameter set; The dynamic modeling module calculates the daily change rate of grain pile temperature and the gradient change rate of weekly fluctuation amplitude of humidity gradient based on the core parameter set, divides the parameters into key layers based on sensitivity, adopts short-term dynamic adjustment method, and adopts trend fitting method for auxiliary layers, to generate a hierarchical model parameter set; The anomaly detection module calls the core parameter set and the hierarchical model parameter set, monitors the residual absolute value of the measured value and the predicted value of the grain pile temperature, and determines whether it exceeds the dynamic anomaly threshold. If the residual exceeds the threshold continuously, an anomaly mark is triggered to generate an environmental anomaly index value; The optimization configuration module adjusts the calculation frequency of the differentiated sensitivity model parameters based on the environmental anomaly index value, combines the gas concentration priority in the ventilation stage and the temperature difference weight in the static stage, reconfigures the parameter screening threshold, and generates an optimized parameter configuration plan.

[0007] As a further scheme of the present invention, the multi-parameter standardized observation values ​​include standardized values ​​of grain pile temperature, standardized values ​​of grain surface humidity gradient, and standardized values ​​of gas concentration parameters; the core parameter set includes grain pile temperature, grain surface humidity gradient, and gas concentration parameters; the layered model parameter set includes key layer parameters and auxiliary layer parameters; the environmental abnormality index values ​​include grain pile temperature abnormality indicators, humidity gradient abnormality indicators, and gas concentration abnormality indicators.

[0008] As a further solution of the present invention, the environmental parameter monitoring module includes: The data acquisition submodule obtains the grain pile temperature, grain surface humidity gradient, and gas concentration parameters in the storage environment. The temperature sensor, humidity sensor, and gas sensor collect the data in real time, and timestamp the raw data of multiple sensors to establish the original environmental data set. The data processing submodule calculates the moving average of the time series based on the environmental original data set and the sensor data of different types, and smoothes the temperature data, humidity data, and gas concentration data respectively, and screens out abnormal data points to establish a sliding average data set; The standardization calculation submodule calls the sliding average data set to normalize the grain pile temperature, grain surface humidity gradient, and gas concentration parameters using the formula: ; Obtain standardized observation values ​​of grain pile temperature, grain surface humidity gradient, and gas concentration through calculation, and generate multi-parameter standardized observation values; in, represents a standardized observed value of a single environmental parameter, Represents the parameter value of the current sampling point, , are the minimum and maximum values ​​of the parameters in the data set, is the sliding mean of the parameter, is the standard deviation of the parameter.

[0009] As a further solution of the present invention, the parameter screening module includes: The joint probability calculation submodule calculates the joint probability distribution of the grain pile temperature and the grain surface humidity gradient based on the multi-parameter standardized observation values, obtains the marginal probability density of the temperature and humidity gradient, constructs a joint probability distribution matrix, and normalizes the matrix to ensure that the total probability is 1, thereby obtaining the joint probability distribution of the temperature and humidity gradient; The mutual information calculation submodule calculates the mutual information value of the gas concentration parameter and the target variable based on the temperature and humidity gradient joint probability distribution, using the formula: ; A set of mutual information values ​​of gas concentrations is calculated; in, represents the adjusted mutual information, At a given gas concentration Temperature under the condition The conditional probability of is the marginal probability of temperature, is the number of gas species, is the number of temperature values; The high correlation coefficient parameter screening submodule compares the mutual information value with a set threshold value based on the gas concentration mutual information value set, screens parameters with mutual information values ​​higher than the threshold value, and establishes a core parameter set.

[0010] As a further solution of the present invention, the dynamic modeling module includes: The temperature and humidity change calculation submodule calls the grain pile temperature and humidity data based on the core parameter set, calculates the daily change rate of the grain pile temperature and the weekly fluctuation range of the humidity gradient, compares the temperature and humidity change characteristics of the differentiated areas inside the grain pile according to the absolute value of the temperature change rate and the change range of the humidity gradient, calculates the temperature and humidity change ratio, analyzes its change trend in multiple layers, and obtains the temperature and humidity change ratio distribution; The gradient change analysis submodule calls the temperature and humidity change ratio distribution, divides the critical levels of the parameters based on sensitivity, extracts the key layer parameters and the auxiliary layer parameters respectively, calculates the short-term dynamic adjustment gradient of the key layer parameters, and obtains the trend fitting error of the auxiliary layer parameters using the formula: ; Calculate the sensitivity of gradient changes and classify them based on the parameter level to obtain the temperature and humidity adjustment gradient of the key layer and the temperature and humidity trend error of the auxiliary layer; in, represents the sensitivity of gradient change, represents the rate of temperature change, Represents the humidity change range, is the adjustment factor determined based on past data. Represents the standard deviation of temperature and humidity data, Represents the temperature values ​​of multiple measuring points in the grain pile, Represents the humidity values ​​at multiple measuring points in the grain pile; The model level parameter optimization submodule calls the key layer temperature and humidity adjustment gradient and the auxiliary layer temperature and humidity trend error, corrects the key layer parameters according to the short-term dynamic adjustment method, and optimizes the auxiliary layer parameters based on the trend fitting results, integrating the two types of level parameters to establish a hierarchical model parameter set.

[0011] As a further solution of the present invention, the anomaly detection module includes: The threshold calculation submodule calls the core parameter set and the hierarchical model parameter set, based on volatility adjustment and seasonal factor correction, using the formula: ; The dynamic abnormal threshold is calculated; in, Representative time Dynamic abnormal threshold at the moment, Representative time The mean predicted value at time, Representative time The standard deviation of the predicted value at time, represents the volatility correction factor, represents the prediction error amplification factor, represents the global regulatory factor; The residual analysis submodule calculates the residual absolute value of the actual temperature measurement value and the predicted value of the grain pile based on the dynamic abnormal threshold, sets the comparison parameters, and calculates the corresponding exceeded interval based on the difference. If the dynamic abnormal threshold is exceeded continuously, the interval is determined to be an abnormal fluctuation interval. The abnormal trigger and cumulative analysis submodule calls the abnormal fluctuation interval, sets the trigger threshold, calculates the abnormal cumulative value according to the duration and intensity of the abnormal interval, compares the abnormal trigger standard, and generates the environmental abnormal index value if the abnormal conditions are met.

[0012] As a further solution of the present invention, the system further includes: The optimization configuration module adjusts the calculation frequency of the differential sensitivity model parameters based on the environmental abnormality index value, combines the gas concentration priority in the ventilation stage and the temperature difference weight in the static stage, reconfigures the parameter screening threshold, and generates an optimization parameter configuration scheme; The optimization parameter configuration scheme specifically includes a calculation frequency adjustment scheme, a parameter screening threshold reconfiguration scheme, and an optimization weight scheme for the ventilation stage and the static stage.

[0013] As a further solution of the present invention, the optimization configuration module includes: The abnormal index calculation submodule obtains the gas concentration in the ventilation stage and the temperature difference data in the static stage based on the abnormal environmental index value, calculates the mean, variance and change rate, constructs a differentiated sensitivity factor, compares the data fluctuations in the differentiated time period, and obtains a differentiated sensitivity factor matrix; The sensitivity model adjustment submodule calls the differentiated sensitivity factor matrix, adjusts the calculation frequency according to the gas concentration priority in the ventilation stage and the temperature difference weight in the static stage, and adopts the formula: ; Adjustment is performed in combination with the gas concentration priority to obtain the optimized calculation frequency parameters; in, represents the optimization calculation frequency parameter, represents the differential sensitivity factor, represents the temperature difference weight during the static stage, Represents the fluctuation amplitude of gas concentration, Represents the current temperature, represents the base temperature; The parameter screening and optimization submodule calls the optimization calculation frequency parameters, screens the parameter combination with larger gas concentration weight in the ventilation stage, compares the screening results under the differentiated screening threshold, establishes differentiated sensitivity parameter screening rules, and obtains the optimization parameter configuration scheme.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, key environmental parameters are screened by combining probability distribution and mutual information calculation methods, factors affecting grain storage quality are accurately identified, and the pertinence of risk assessment is enhanced. Hierarchical dynamic modeling is adopted, and parameters at different levels are respectively adjusted dynamically in a short time and trend fitting is performed, so that the prediction can not only respond quickly to short-term changes, but also reflect long-term trends. Anomaly detection improves the accuracy of abnormal warning through residual analysis of measured values ​​and predicted values, combined with dynamic threshold adjustment. The optimization configuration link adjusts the calculation frequency in combination with the characteristics of the storage stage, and reallocates the parameter screening weights based on different environmental states to improve the control accuracy and reduce energy consumption. The overall solution improves the accuracy of storage management in terms of environmental monitoring, anomaly detection, and optimized configuration, making grain storage safer and storage losses lower. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a system flow chart of the present invention; Figure 2 This is a flow chart of the environmental parameter monitoring module of the present invention; Figure 3 This is a flow chart of the parameter screening module of the present invention; Figure 4 This is a flow chart of the dynamic modeling module of the present invention; Figure 5 This is a flow chart of the abnormality detection module of the present invention; Figure 6 This is a flow chart of the optimized configuration module of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0018] Embodiment 1: See also Figure 1 , a grain storage management system based on Internet of Things technology includes: The environmental parameter monitoring module collects grain pile temperature, grain surface humidity gradient, and gas concentration parameters in the storage environment in real time through sensors, performs sliding average calculation and standardization on the data, and generates multi-parameter standardized observation values; The parameter screening module calls the standardized observation values ​​of multiple parameters, calculates the joint probability distribution of grain pile temperature and grain surface humidity gradient, calculates the mutual information value of gas concentration parameter and target variable, and screens the parameters with correlation with grain pile mildew risk higher than the threshold by comparing the mutual information value with the set threshold, thus generating a core parameter set. The dynamic modeling module calculates the daily change rate of grain pile temperature and the gradient change rate of weekly fluctuation amplitude of humidity gradient based on the core parameter set, divides the parameters into key layers based on sensitivity, adopts short-term dynamic adjustment method, and adopts trend fitting method for auxiliary layers, to generate a hierarchical model parameter set; The anomaly detection module calls the core parameter set and the hierarchical model parameter set to monitor the absolute value of the residual between the measured and predicted values ​​of the grain pile temperature and determine whether it exceeds the dynamic anomaly threshold. If the residual exceeds the threshold continuously, the anomaly mark is triggered and the environmental anomaly index value is generated. The optimization configuration module adjusts the calculation frequency of the differentiated sensitivity model parameters based on the environmental anomaly index value, combines the gas concentration priority in the ventilation stage and the temperature difference weight in the static stage, reconfigures the parameter screening threshold, and generates an optimized parameter configuration plan.

[0019] The multi-parameter standardized observation values ​​include the standardized values ​​of grain pile temperature, the standardized values ​​of grain surface humidity gradient, and the standardized values ​​of gas concentration parameters. The core parameter set includes grain pile temperature, grain surface humidity gradient, and gas concentration parameters. The layered model parameter set includes key layer parameters and auxiliary layer parameters. The environmental anomaly index values ​​include grain pile temperature anomaly index, humidity gradient anomaly index, and gas concentration anomaly index. The optimized parameter configuration scheme specifically includes the calculation frequency adjustment scheme, the parameter screening threshold reconfiguration scheme, and the optimized weight scheme for the ventilation stage and the static stage.

[0020] See also Figure 2 , the environmental parameter monitoring module includes: The data acquisition submodule obtains the grain pile temperature, grain surface humidity gradient, and gas concentration parameters in the storage environment. The temperature sensor, humidity sensor, and gas sensor collect the data in real time, and timestamp the raw data of multiple sensors to establish the original environmental data set. First, multiple measurement points are arranged at different heights and locations of the granary. For example, a measurement point is set every 2 meters, and data is obtained through temperature and humidity probes at different depths. For gas concentration parameters, sensors are arranged above, in the middle and at the bottom of the granary to form concentration monitoring in the spatial dimension. All sensors need to be triggered synchronously at each measurement moment and record the corresponding timestamp. The timestamp uses the UNIX time format. For example, "1709876543" represents the time point of a certain data collection. The value read by the temperature sensor is recorded in degrees Celsius. Assuming that the temperature data of a certain measurement point is 18.4℃, the humidity data is 67.5%, and the carbon dioxide concentration data is 0.04% (400ppm), the raw data collected by the sensor is stored in the database and classified and stored according to the measurement point number. For example, the data storage format of a certain measurement point is (1709876543, 18.4, 67.5, 0.04%). The original environmental data set constructed in this way covers the data of different measurement points and can be used for subsequent analysis.

[0021] The data processing submodule calculates the moving average of the time series based on the original environmental data set and the sensor data of different types, and smoothes the temperature data, humidity data, and gas concentration data respectively, and screens out abnormal data points to establish a sliding average data set; First, the data of each measurement point are arranged in time series, and the set window size is selected to calculate the moving average. For example, the window size is set to 5, that is, the arithmetic average of the past 5 data points is taken. If the temperature data of a certain measurement point in the last 5 measurements are 18.4℃, 18.2℃, 18.5℃, 18.3℃, and 18.7℃, the moving average is calculated as (18.4+18.2+18.5+18.3+18.7) / 5=18.42℃. The humidity and gas concentration data are also calculated in the same way. The data is smoothed by the formula. Next, outliers are screened out and the standard deviation range method is used to calculate the mean and standard deviation of the measurement point data. Assuming that the mean temperature of the measurement point is 18.42℃ and the standard deviation is 0.2℃, if the temperature data of a certain measurement exceeds the mean ± 3 times the standard deviation (i.e., the range of 18.42±0.6℃), for example, a temperature value of 19.5℃ appears, the data is judged as an outlier and is removed. The humidity and gas concentration data use the same outlier removal method to finally generate a sliding average data set.

[0022] The standardization calculation submodule calls the sliding average data set to normalize the grain pile temperature, grain surface humidity gradient, and gas concentration parameters using the formula: ; Obtain standardized observation values ​​of grain pile temperature, grain surface humidity gradient, and gas concentration through calculation, and generate multi-parameter standardized observation values; in, represents a standardized observed value of a single environmental parameter, Represents the parameter value of the current sampling point, , are the minimum and maximum values ​​of the parameters in the data set, is the sliding mean of the parameter, is the standard deviation of the parameter.

[0023] First, calculate the minimum and maximum values ​​in the data set. For example, in the temperature data set of a certain measuring point, the minimum temperature value is 16.5℃ and the maximum temperature value is 19.8℃. Then, calculate the sliding average value. and standard deviation , assuming that in a certain measurement, the temperature data , mean , standard deviation , then according to the normalization formula: ; Substitute the parameter values ​​into: ; ; Similarly, for humidity data, assume that the minimum value is 60% and the maximum value is 80%. , , then if the humidity value at a certain measuring point is , calculate the standardized value: ; The results show that the standardized value of temperature at this measuring point is 0.676, and the standardized value of humidity is 0.8, both of which are within the standardized range [0,1], ensuring the comparison of data from different measuring points at the same scale. This value can be used for subsequent data modeling or anomaly detection. When further sorted, the standardized data of all measuring points can be calculated and stored in the database for subsequent analysis, as shown in Table 1.

[0024] Table 1 Standardized observation data set:

[0025] As shown in Table 1, the standardized data of each measuring point has been calculated and can be used for further analysis and processing. This result shows that the data of each measuring point has been converted to a standard scale through normalization calculation, so that the data of different environmental parameters can be directly compared, which is convenient for subsequent data trend analysis, anomaly detection and the establishment of prediction models.

[0026] See also Figure 3 , the parameter screening module includes: The joint probability calculation submodule calculates the joint probability distribution of the grain pile temperature and the grain surface humidity gradient based on the standardized observation values ​​of multiple parameters, obtains the marginal probability density of the temperature and humidity gradient, constructs the joint probability distribution matrix, and normalizes the matrix to ensure that the total probability is 1, thus obtaining the joint probability distribution of the temperature and humidity gradient. First, the temperature of the grain pile Grain surface moisture gradient Standardize to normalize the value range to , assuming that the grain pile temperature range is to The humidity gradient ranges from to The standardized calculation is as follows: ; ; For example, when the temperature of a measuring point is When , the standardized temperature is: ; When the humidity gradient is measured When , the standardized humidity gradient is: ; Next, calculate the temperature and humidity gradient The joint probability is the frequency of occurrence of different temperature and humidity combinations in all observed samples. For example, suppose there are 1000 data points, among which the temperature and humidity combinations It appears 40 times, then the joint probability of the combination is calculated as follows: ; Then calculate the marginal probability density, that is, sum the temperature and humidity gradients separately, for example: ; ; Finally, all joint probability values ​​are formed into a matrix and normalized to satisfy: ; That is, the sum of all probabilities must be equal to 1 to ensure the rationality of the joint probability distribution, and finally obtain the joint probability distribution matrix of the temperature and humidity gradient.

[0027] The mutual information calculation submodule calculates the mutual information value of the gas concentration parameter and the target variable based on the temperature and humidity gradient joint probability distribution, using the formula: ; A set of mutual information values ​​of gas concentrations is calculated; in, represents the adjusted mutual information, At a given gas concentration Temperature under the condition The conditional probability of is the marginal probability of temperature, is the number of gas species, is the number of temperature values; The specific implementation process is as follows: Step 1: Calculate the conditional probability of temperature: For a certain gas concentration , calculate the temperature The conditional probability under the gas concentration condition is , calculated as follows: ; For example, assuming that carbon dioxide ( ) concentration is 0.2 in a certain range, and the temperature under this gas concentration The joint probability of , and the gas concentration The marginal probability is , then calculate: ; Step 2: Calculate the marginal probability of temperature: For temperature , calculate its marginal probability: ; Assumptions , , ,but: ; Step 3: Calculate the mutual information value: Using the formula: ; For each gas concentration And all temperature values , calculate the information entropy transformation term: ; For example, it is known that , ,but: ; Assumptions , ,but: ; Finally, the mutual information value of the gas is obtained by calculating the weighted sum of the mutual information values ​​of all temperature values.

[0028] The high correlation coefficient parameter screening submodule is based on the gas concentration mutual information value set, compares the mutual information value with the set threshold, screens the parameters with mutual information values ​​higher than the threshold, and establishes the core parameter set.

[0029] Step 1: Set the screening threshold: Setting Thresholds To filter out high-related gas concentration parameters, for example, thresholds can be set .

[0030] Step 2: Traverse the mutual information value set: The calculated gas concentration mutual information value set is as follows: ; Step 3: Screening core parameters: Compare the mutual information value with the set threshold , filter out gas parameters that meet the conditions: (no filtering), (filter), (filter), (no filtering), (filter); Finally, the core parameter set is obtained: ; Table 2 Gas concentration mutual information calculation results:

[0031] As shown in Table 2, the gas parameters with high mutual information values ​​finally screened out are , the mutual information values ​​of these gases all exceed the set threshold , and therefore is included in the core parameter set.

[0032] See also Figure 4 , dynamic modeling modules include: The temperature and humidity change calculation submodule calls the grain pile temperature and humidity data based on the core parameter set, calculates the daily change rate of grain pile temperature and the weekly fluctuation range of humidity gradient, compares the temperature and humidity change characteristics of differentiated areas inside the grain pile according to the absolute value of the temperature change rate and the change range of the humidity gradient, calculates the temperature and humidity change ratio, analyzes its change trend in multiple layers, and obtains the temperature and humidity change ratio distribution; First, obtain the temperature and humidity data of multiple measuring points inside the grain pile. The data is recorded regularly by sensors. The sampling interval is usually set to 1 hour. The number of measuring points is determined according to the size of the grain silo. For example, a grain pile of 10m×10m×5m can be equipped with 36 measuring points. The data includes temperature, humidity, and humidity. (Unit: °C) and humidity (Unit: %RH).

[0033] When calculating the temperature change rate, the time interval is set to 24 hours, and the calculation formula is as follows: ; Taking actual data as an example, assuming that the temperature at a certain measuring point rises from 25.2°C to 27.1°C within 24 hours, the daily change rate is: ; The weekly fluctuation amplitude of humidity gradient is calculated using 7-day period data and is defined as: ; If the maximum humidity at a certain measuring point is 72.5%RH and the minimum is 65.8%RH within 7 days, then: ; The next step is to calculate the temperature and humidity change ratio. It is necessary to compare the temperature and humidity change characteristics of different areas and define the ratio calculation method: ; Bring in the above data: ; The results show that at this measuring point, the hourly rate of change of temperature is lower than the fluctuation amplitude of humidity, and the ratio of temperature to humidity is 0.0118, indicating that the temperature and humidity changes in this area are relatively stable. If the ratio is generally small in the entire grain pile, it means that the environmental changes inside the grain pile are relatively uniform; on the contrary, if the ratios in different areas differ greatly, there may be areas of temperature and humidity imbalance, which require further adjustment or monitoring.

[0034] After calculating the ratios of all measuring points, they are divided into levels and summarized to form the distribution of temperature and humidity change ratios. The data example is as follows: Table 3 Temperature and humidity change ratio distribution table:

[0035] As shown in Table 3, the temperature and humidity change ratios of each measuring point are concentrated between 0.011 and 0.012. This value can be used to determine the stability of the temperature and humidity changes inside the grain pile and provide input data for subsequent gradient change analysis.

[0036] The gradient change analysis submodule calls the temperature and humidity change ratio distribution, divides the critical levels of parameters based on sensitivity, extracts the key layer parameters and auxiliary layer parameters respectively, calculates the short-term dynamic adjustment gradient of the key layer parameters, and obtains the trend fitting error of the auxiliary layer parameters using the formula: ; Calculate the sensitivity of gradient changes and classify them based on the parameter level to obtain the temperature and humidity adjustment gradient of the key layer and the temperature and humidity trend error of the auxiliary layer; in, represents the sensitivity of gradient change, represents the rate of temperature change, Represents the humidity change range, is the adjustment factor determined based on past data. Represents the standard deviation of temperature and humidity data, Represents the temperature values ​​of multiple measuring points in the grain pile, Represents the humidity values ​​at multiple measuring points in the grain pile; First, define the sensitivity threshold, that is, when the ratio Above a certain threshold When the threshold is 0.0115, the key layer points include points 1, 2, and 3, while point 4 is classified as an auxiliary layer.

[0037] The key layer short-term dynamic adjustment gradient calculation is as follows: ; in: Derived from historical data fitting, assumed to be 0.0085- is the standard deviation of temperature and humidity, calculated as: ; in °C, , the standard deviation is calculated .

[0038] Substitute the value of measurement point 1: ; The result shows that the temperature and humidity adjustment gradient of measuring point 1 is small, which means that its temperature and humidity parameters are relatively stable. If the value is large (for example, higher than 0.02), it may indicate that there are strong temperature and humidity fluctuations in the area, and the key layer parameters need to be further adjusted.

[0039] The model-level parameter optimization submodule calls the temperature and humidity adjustment gradient of the key layer and the temperature and humidity trend error of the auxiliary layer, corrects the key layer parameters according to the short-term dynamic adjustment method, and optimizes the auxiliary layer parameters based on the trend fitting results, integrating the two types of level parameters to establish a hierarchical model parameter set.

[0040] First, adjust the key layer parameters using dynamic weight adjustment: ; in: To adjust the weight, assume it is 0.6, is the initial parameter, assuming that the initial parameter value of measuring point 1 is 5.2; but: ; The result shows that the adjustment value of the key layer measuring point has a small increment (0.0076), which reflects that the temperature and humidity stability of the measuring point is relatively high and the required correction amplitude is relatively small. If the increment is large (for example, higher than 0.05), it indicates that the temperature and humidity changes are unstable and require a larger adjustment.

[0041] For the auxiliary layer, the trend fitting results are used to correct the parameters: ; in: is the trend error. Assuming the error of measuring point 4 is 0.025, is the correction coefficient, take 0.7; but: ; The results show that the parameter correction amount of the auxiliary layer measuring point is small, which means that its temperature and humidity change trend basically conforms to the prediction model. If the correction amount is large (such as higher than 0.1), the trend fitting model may need to be readjusted.

[0042] See also Figure 5 , the anomaly detection module includes: The threshold calculation submodule calls the core parameter set and the hierarchical model parameter set, based on volatility adjustment and seasonal factor correction, using the formula: ; The dynamic abnormal threshold is calculated; in, Representative time Dynamic abnormal threshold at the moment, Representative time The mean predicted value at time, Representative time The standard deviation of the predicted value at time, represents the volatility correction factor, represents the prediction error amplification factor, represents the global regulatory factor; formula: ; First, the predicted value of the grain pile temperature is processed to extract the time The predicted mean at time With standard deviation During the calculation process, the system predicts the temperature at a daily or hourly granularity based on the historical temperature data set of the grain warehouse, and obtains the mean and standard deviation at the same time point. For example, in the historical data of a certain storage point, the mean temperature of the grain pile at 08:00 on a certain day is is 28.5℃, standard deviation is 2.1℃, then the system calls the volatility correction factor , which is calculated based on the recent temperature change rate. Assuming the maximum temperature difference in the past 7 days is 6°C, then Can be set to 1.15 (experience range 1.0-1.2), and the prediction error magnification factor is called at the same time , the coefficient is calculated by normalizing the historical error variance. Assuming that the mean square error of the forecast error in the historical data is 1.8℃, after normalization The value is 0.85, and finally the global adjustment factor is called , this factor is used to adjust the overall level of the threshold. Assuming it is set to 1.05, the final dynamic anomaly threshold is calculated as follows: ; ; ; The results show that the dynamic abnormal threshold at the current moment has been calculated, and based on the correction of volatility, prediction error and global adjustment factor, the threshold can dynamically adapt to the temperature change trend. The numerical result will be used in the subsequent residual calculation step to determine whether the grain pile temperature has entered the abnormal range.

[0043] The residual analysis submodule calculates the residual absolute value of the measured value and the predicted value of the grain pile temperature based on the dynamic abnormal threshold, sets the comparison parameters, and calculates the corresponding exceeded interval based on the difference. If the dynamic abnormal threshold is exceeded continuously, the interval is determined to be an abnormal fluctuation interval. Based on the dynamic abnormal threshold, the residual absolute value of the measured value and the predicted value of the grain pile temperature is calculated, that is: ; in is the measured temperature. Assuming that at 08:00, the measured temperature at a certain measuring point is 34.2°C, then calculate: ; Next, set the comparison parameters. Beyond If the calculated dynamic abnormal threshold is exceeded, the moment is judged as an abnormal point. If the time exceeding the threshold continuously reaches the set time (for example, 4 hours), the time interval is judged as an abnormal fluctuation interval. Assuming that the temperature of the measuring point exceeds the calculated dynamic abnormal threshold for 4 consecutive hours, the abnormal fluctuation interval is confirmed to be 08:00 to 12:00. This result shows that the residual of the current measuring point has exceeded the normal range, and the continuous abnormal time has reached the set threshold. Therefore, it is necessary to enter the abnormal trigger and cumulative analysis step to further evaluate the severity of the abnormality.

[0044] The anomaly triggering and cumulative analysis submodule calls the abnormal fluctuation interval, sets the trigger threshold, calculates the abnormal cumulative value according to the duration and intensity of the abnormal interval, and compares it with the abnormal trigger standard. If the abnormal conditions are met, the environmental abnormality indicator value is generated.

[0045] Assume that the trigger threshold is set to 3 hours, that is, the abnormal calculation is triggered when the abnormal interval lasts for 3 hours or more, then 08:00-12:00 meets the trigger condition, and then the abnormal cumulative value is calculated. The calculation method of the abnormal cumulative value is: ; Assume that the measured temperatures from 08:00 to 12:00 are 34.2°C, 35.1°C, 36.3°C, and 37.0°C respectively, and substitute them into the calculation: ; ; Compare the abnormal trigger standard, assuming that the abnormal trigger standard is set to ,but The abnormal triggering standard has not been reached, so the current accumulated abnormal value is not enough to trigger the environmental abnormal index value, and it is necessary to continue to monitor the subsequent time data. This result shows that although the abnormal phenomenon has occurred, the current abnormal accumulation has not reached the threshold set by the system, so it will not immediately trigger a higher level of abnormal alarm, but it is necessary to continue to monitor the subsequent data to determine whether to further upgrade the abnormal state.

[0046] Table 4 Monitoring point temperature data table:

[0047] As shown in Table 4, the measured temperatures at the monitoring points all exceeded the dynamic anomaly threshold for 4 hours, meeting the definition of the abnormal fluctuation range. However, the abnormal accumulation value has not yet triggered the calculation of the environmental anomaly index. This result shows that the current abnormal accumulation level is low and has not yet reached the condition for triggering an alarm. Further observation of subsequent temperature change trends is needed to decide whether to take further measures.

[0048] See also Figure 6 , the optimization configuration modules include: The abnormal index calculation submodule obtains the gas concentration in the ventilation stage and the temperature difference data in the static stage based on the environmental abnormal index value, calculates the mean, variance and change rate, constructs the differentiated sensitivity factor, compares the data fluctuations in the differentiated time period, and obtains the differentiated sensitivity factor matrix; First, obtain the gas concentration data during the ventilation stage and the temperature difference data during the static stage. Data acquisition can be recorded in real time by sensors installed at different monitoring points. Assume that the gas concentration data (unit: ppm) of a certain monitoring point during the ventilation stage is , the temperature difference data (unit: ℃) during the static stage is Next, calculate the mean, variance, and rate of change of these data. The mean is calculated as ,in represents the data value, is the number of data, for example, the mean gas concentration is ; The mean temperature difference is ℃; The formula for calculating variance is: ; The variance of the calculated gas concentration is ; The variance of the temperature difference is ; The rate of change is obtained by calculating the average value of adjacent data changes. For example, the rate of change of gas concentration is ppm / time; The rate of change of static temperature difference is ℃ / time; Next, we construct a differential sensitivity factor, defined as , the differential sensitivity factor of gas concentration is calculated as , the differential sensitivity factor of temperature difference is ,Finally, the data fluctuations in different time periods are compared to form a ,differentiated sensitivity factor matrix, as shown in Table 5.

[0049] Table 5 Differentiation sensitivity factor matrix:

[0050] As shown in Table 5, the sensitivity factor of gas concentration is higher than that of temperature difference, indicating that the gas concentration fluctuates more.

[0051] The sensitivity model adjustment submodule calls the differentiated sensitivity factor matrix, adjusts the calculation frequency according to the gas concentration priority in the ventilation stage and the temperature difference weight in the static stage, and uses the formula: ; Adjustment is performed in combination with the gas concentration priority to obtain the optimized calculation frequency parameters; in, represents the optimization calculation frequency parameter, represents the differential sensitivity factor, represents the temperature difference weight during the static stage, Represents the fluctuation amplitude of gas concentration, Represents the current temperature, represents the base temperature; First, call the differentiated sensitivity factor matrix, and use the gas concentration priority and the temperature difference weight in the static stage as the basis for adjustment. Assuming that the gas concentration priority is 0.7 and the temperature difference weight is 0.3, calculate the optimized calculation frequency parameters , during the calculation process, Calculated as , When calculating, assume that the gas concentration fluctuation range , current temperature ℃, reference temperature ℃, calculated , substitute into the formula: ; Then, according to the calculation results and the gas concentration priority, adjust and set the optimized calculation frequency The threshold interval is 0.03-0.06, and the current calculation result 0.0515 falls within this interval, so it is considered that the optimization calculation frequency parameter is within a reasonable range.

[0052] The parameter screening and optimization submodule calls the optimized calculation frequency parameters, screens the parameter combinations with larger gas concentration weights in the ventilation stage, compares the screening results under the differentiated screening thresholds, establishes differentiated sensitivity parameter screening rules, and obtains the optimized parameter configuration plan.

[0053] Assuming that there are five sets of data combinations, calculate their gas concentration weights and optimize the calculation frequency parameters, as shown in Table 6: Table 6 Parameter screening and optimization results:

[0054] As shown in Table 6, the screening threshold is set to 0.05, the parameter combinations A and B that meet the threshold are screened out, and the differentiated sensitivity parameter screening rules are established, and finally the optimized parameter configuration scheme is obtained.

[0055] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A grain storage management system based on Internet of Things technology, characterized in that: The system comprises: The environmental parameter monitoring module collects grain pile temperature, grain surface humidity gradient, and gas concentration parameters in the storage environment in real time through sensors, performs sliding average calculation and standardization on the data, and generates multi-parameter standardized observation values; The parameter screening module calls the multi-parameter standardized observation value, calculates the joint probability distribution of the grain pile temperature and the grain surface humidity gradient, calculates the mutual information value of the gas concentration parameter and the target variable, and screens the parameters with a correlation with the risk of mildew in the grain pile higher than the threshold by comparing the mutual information value with the set threshold, thereby generating a core parameter set; The dynamic modeling module calculates the daily change rate of grain pile temperature and the gradient change rate of weekly fluctuation amplitude of humidity gradient based on the core parameter set, divides the parameters into key layers based on sensitivity, adopts short-term dynamic adjustment method, and adopts trend fitting method for auxiliary layers, to generate a hierarchical model parameter set; The anomaly detection module calls the core parameter set and the hierarchical model parameter set, monitors the absolute value of the residual between the measured value and the predicted value of the grain pile temperature, and determines whether it exceeds the dynamic anomaly threshold. If the residual exceeds the threshold continuously, the anomaly mark is triggered and the environmental anomaly index value is generated.

2. The food storage management system based on Internet of Things technology according to claim 1 is characterized in that: The multi-parameter standardized observation values ​​include standardized values ​​of grain pile temperature, standardized values ​​of grain surface humidity gradient, and standardized values ​​of gas concentration parameters. The core parameter set includes grain pile temperature, grain surface humidity gradient, and gas concentration parameters. The layered model parameter set includes key layer parameters and auxiliary layer parameters. The environmental anomaly index values ​​include grain pile temperature anomaly index, humidity gradient anomaly index, and gas concentration anomaly index.

3. The food storage management system based on Internet of Things technology according to claim 2 is characterized in that: The environmental parameter monitoring module includes: The data acquisition submodule obtains the grain pile temperature, grain surface humidity gradient, and gas concentration parameters in the storage environment. The temperature sensor, humidity sensor, and gas sensor collect the data in real time, and timestamp the raw data of multiple sensors to establish the original environmental data set. The data processing submodule calculates the moving average of the time series based on the environmental original data set and the sensor data of different types, and smoothes the temperature data, humidity data, and gas concentration data respectively, and screens out abnormal data points to establish a sliding average data set; The standardization calculation submodule calls the sliding average data set to normalize the grain pile temperature, grain surface humidity gradient, and gas concentration parameters using the formula: ； Obtain standardized observation values ​​of grain pile temperature, grain surface humidity gradient, and gas concentration through calculation, and generate multi-parameter standardized observation values; in, represents a standardized observed value of a single environmental parameter, Represents the parameter value of the current sampling point, , are the minimum and maximum values ​​of the parameters in the data set, is the sliding mean of the parameter, is the standard deviation of the parameter.

4. The food storage management system based on Internet of Things technology according to claim 3 is characterized in that: The parameter screening module includes: The joint probability calculation submodule calculates the joint probability distribution of the grain pile temperature and the grain surface humidity gradient based on the multi-parameter standardized observation values, obtains the marginal probability density of the temperature and humidity gradient, constructs a joint probability distribution matrix, and normalizes the matrix to ensure that the total probability is 1, thereby obtaining the joint probability distribution of the temperature and humidity gradient; The mutual information calculation submodule calculates the mutual information value of the gas concentration parameter and the target variable based on the temperature and humidity gradient joint probability distribution, using the formula: ； A set of mutual information values ​​of gas concentrations is calculated; in, represents the adjusted mutual information, At a given gas concentration Temperature under the condition The conditional probability of is the marginal probability of temperature, is the number of gas species, is the number of temperature values; The high correlation coefficient parameter screening submodule compares the mutual information value with a set threshold value based on the gas concentration mutual information value set, screens parameters with mutual information values ​​higher than the threshold value, and establishes a core parameter set.

5. The food storage management system based on Internet of Things technology according to claim 4 is characterized in that: The dynamic modeling module includes: The temperature and humidity change calculation submodule calls the grain pile temperature and humidity data based on the core parameter set, calculates the daily change rate of the grain pile temperature and the weekly fluctuation range of the humidity gradient, compares the temperature and humidity change characteristics of the differentiated areas inside the grain pile according to the absolute value of the temperature change rate and the change range of the humidity gradient, calculates the temperature and humidity change ratio, analyzes its change trend in multiple layers, and obtains the temperature and humidity change ratio distribution; The gradient change analysis submodule calls the temperature and humidity change ratio distribution, divides the critical levels of parameters based on sensitivity, extracts the key layer parameters and the auxiliary layer parameters respectively, calculates the short-term dynamic adjustment gradient of the key layer parameters, and obtains the trend fitting error of the auxiliary layer parameters using the formula: ； Calculate the sensitivity of gradient changes and classify them based on the parameter level to obtain the temperature and humidity adjustment gradient of the key layer and the temperature and humidity trend error of the auxiliary layer; in, represents the sensitivity of gradient change, represents the rate of temperature change, Represents the humidity change range, is the adjustment factor determined based on past data. Represents the standard deviation of temperature and humidity data, Represents the temperature values ​​of multiple measuring points in the grain pile, Represents the humidity values ​​at multiple measuring points in the grain pile; The model level parameter optimization submodule calls the key layer temperature and humidity adjustment gradient and the auxiliary layer temperature and humidity trend error, corrects the key layer parameters according to the short-term dynamic adjustment method, and optimizes the auxiliary layer parameters based on the trend fitting results, integrating the two types of level parameters to establish a hierarchical model parameter set.

6. The food storage management system based on Internet of Things technology according to claim 5 is characterized in that: The anomaly detection module comprises: The threshold calculation submodule calls the core parameter set and the hierarchical model parameter set, based on volatility adjustment and seasonal factor correction, using the formula: ； The dynamic abnormal threshold is calculated; in, Representative time Dynamic abnormal threshold at the moment, Representative time The mean predicted value at time, Representative time The standard deviation of the predicted value at time, represents the volatility correction factor, represents the prediction error amplification factor, represents the global regulatory factor; The residual analysis submodule calculates the residual absolute value of the actual measured value and the predicted value of the grain pile temperature based on the dynamic abnormal threshold, sets the comparison parameters, and calculates the corresponding exceeded interval based on the difference. If the dynamic abnormal threshold is exceeded continuously, the interval is determined to be an abnormal fluctuation interval. The abnormal trigger and cumulative analysis submodule calls the abnormal fluctuation interval, sets the trigger threshold, calculates the abnormal cumulative value according to the duration and intensity of the abnormal interval, compares the abnormal trigger standard, and generates the environmental abnormal index value if the abnormal conditions are met.

7. The food storage management system based on Internet of Things technology according to claim 6 is characterized in that: The system further comprises: The optimization configuration module adjusts the calculation frequency of the differential sensitivity model parameters based on the environmental anomaly index value, combines the gas concentration priority in the ventilation stage and the temperature difference weight in the static stage, reconfigures the parameter screening threshold, and generates an optimization parameter configuration scheme; The optimization parameter configuration scheme specifically includes a calculation frequency adjustment scheme, a parameter screening threshold reconfiguration scheme, and an optimization weight scheme for the ventilation stage and the static stage.

8. The food storage management system based on Internet of Things technology according to claim 7 is characterized in that: The optimization configuration module includes: The abnormal index calculation submodule obtains the gas concentration in the ventilation stage and the temperature difference data in the static stage based on the abnormal environmental index value, calculates the mean, variance and change rate, constructs a differentiated sensitivity factor, compares the data fluctuations in the differentiated time period, and obtains a differentiated sensitivity factor matrix; The sensitivity model adjustment submodule calls the differentiated sensitivity factor matrix, adjusts the calculation frequency according to the gas concentration priority in the ventilation stage and the temperature difference weight in the static stage, and adopts the formula: ； Adjustment is performed in combination with the gas concentration priority to obtain the optimized calculation frequency parameters; in, represents the optimization calculation frequency parameter, represents the differential sensitivity factor, represents the temperature difference weight during the static stage, Represents the fluctuation amplitude of gas concentration, Represents the current temperature, represents the base temperature; The parameter screening and optimization submodule calls the optimization calculation frequency parameter, screens the parameter combination of the gas concentration in the ventilation stage, compares the screening results under the differentiated screening threshold, establishes the differentiated sensitivity parameter screening rules, and obtains the optimization parameter configuration scheme.

Images