Intelligent monitoring system and method for refrigeration house environment based on multi-source data fusion

Abstract

The invention discloses a refrigeration house environment intelligent monitoring system and method based on multi-source data fusion, and belongs to the technical field of cold chain monitoring. The system comprises a data acquisition module, a genetic map construction module, a microclimate fingerprint identification module, a thermal inertia compensation module, an intelligent fusion decision module, an abnormality diagnosis and early warning module and a visual display module. Constructing a data gene map by extracting statistical domain, frequency domain and time domain characteristics of multi-source data, and revealing deep association among the data; reconstructing a three-dimensional environment field by adopting Kriging interpolation, and extracting fingerprint features of a microenvironment unit to realize global perception; establishing a heat conduction model to calculate the internal temperature of the cargo, and generating a compensation parameter considering thermal inertia; executing multi-level fusion in the data layer, the feature layer and the decision layer to generate high-credibility comprehensive situation information; and performing anomaly identification and graded early warning based on a fusion result. According to the invention, deep fusion of multi-source heterogeneous data is realized, the monitoring precision and the system reliability are improved, and effective technical support is provided for intelligent management of the refrigeration house.

Description

Technical Field

[0001] This invention belongs to the field of cold chain monitoring technology, specifically relating to an intelligent monitoring system and method for cold storage environment based on multi-source data fusion. Background Technology

[0002] As a crucial link in the cold chain system, the accurate monitoring of environmental parameters in cold storage facilities directly impacts the quality and safety of stored goods. Existing cold storage monitoring technologies have the following main shortcomings: Firstly, traditional monitoring systems often use single-type sensors for independent monitoring, with each sensor's data being independent of the others. This lack of effective data correlation analysis and fusion mechanisms leads to fragmented monitoring information, making it difficult to fully reflect the true state of the cold storage environment.

[0003] Secondly, existing technologies typically employ discrete monitoring methods at fixed points, which cannot detect the complex micro-environmental differences and dynamic changes inside cold storage facilities, resulting in monitoring blind spots and making it easy to miss local anomalies.

[0004] Third, current monitoring systems only focus on environmental parameters themselves, ignoring the differences in the response of different types of goods to temperature changes in their thermophysical properties. This makes it impossible to accurately assess the actual controlled state of the goods, resulting in a mismatch between control strategies and the needs of goods protection.

[0005] Therefore, there is an urgent need for an intelligent monitoring system and method that can achieve deep fusion of multi-source data, full-domain environmental perception, and take into account the characteristics of goods. Summary of the Invention

[0006] To address the technical problems of low data fusion, incomplete environmental perception, and neglect of cargo characteristics in existing cold storage monitoring technologies, this invention provides an intelligent monitoring system and method for cold storage environment based on multi-source data fusion.

[0007] In a first aspect, the present invention provides an intelligent monitoring system for cold storage environment based on multi-source data fusion, the system comprising: The data acquisition module is used to collect environmental monitoring data, equipment operation data and image monitoring data in the cold storage, and to preprocess the raw data. The environmental monitoring data includes temperature data, humidity data, carbon dioxide concentration data and ethylene concentration data. The equipment operation data includes compressor vibration frequency data, operating current data and sound signal data. The image monitoring data includes thermal imaging data and visible light image data. The gene map construction module is used to extract statistical domain, frequency domain and time domain features from the preprocessed data stream as data genes, calculate the mutual information, maximum information coefficient and Granger causality between genes from different data sources, construct a multi-layer association weight matrix containing linear correlation, nonlinear correlation and causal relationship, and form a data gene map that reflects the coupling relationship and evolution law between data sources. The microclimate fingerprinting module is used to discretize the cold storage space into micro-environmental units of preset size. It reconstructs a complete three-dimensional environmental field from sparse sensor data using the Kriging interpolation method. It extracts the mean temperature, temperature gradient magnitude, temperature Laplace operator, airflow direction angle, relative humidity gradient magnitude, temperature-time first derivative, temperature-time second derivative, velocity field divergence, and velocity field curl of each micro-environmental unit as environmental fingerprint features, and establishes a standard fingerprint database and anomaly fingerprint database. The thermal inertia compensation module is used to establish heat conduction equations and lumped parameter models according to cargo type, calculate the equivalent thermophysical parameters of cargo, solve the internal temperature distribution of cargo, establish a dynamic mapping relationship between surface temperature and core temperature, and generate temperature compensation parameters that take into account the thermal response time constant and the cumulative effect of historical temperature of cargo. The intelligent fusion decision module is used to calculate fusion weights based on reputation factors, credibility factors, and relevance factors at the data layer, and perform weighted fusion of similar sensor data; at the feature layer, it splices together gene features, environmental fingerprints, and thermal inertial features, and reduces dimensionality through principal component analysis; at the decision layer, it uses Dempster-Shafer evidence theory to integrate the judgment results of multiple subsystems to generate comprehensive cold storage environment situation information. The anomaly diagnosis and early warning module is used to construct a temperature anomaly pattern vector based on the integrated situational information after fusion, through temperature deviation amplitude, temperature change rate, temperature fluctuation standard deviation, spatial temperature non-uniformity, and anomaly duration; identify equipment faults through the fault characteristic frequency of the vibration spectrum; assess cargo quality risk through the cumulative damage model and Arrhenius equation; and generate graded early warning information using fuzzy logic rules. The visualization module is used to generate a three-dimensional temperature field thermogram using volume rendering technology, map temperature to color through a piecewise linear transfer function, draw time series trend curves of key parameters, and generate an alarm list for abnormal events.

[0008] Furthermore, the gene mapping module employs wavelet transform for multi-scale analysis to extract gene features at different time scales.

[0009] Furthermore, the unit size of the microclimate fingerprint recognition module is dynamically determined based on the ratio of the allowable temperature difference to the maximum temperature gradient.

[0010] Furthermore, the thermal inertia compensation module calculates the equivalent density, equivalent specific heat capacity, and equivalent thermal conductivity for composite material goods using volume weighting.

[0011] Furthermore, the reputation factor of the intelligent fusion decision module is calculated using the exponential decay function of historical prediction errors, the credibility factor is evaluated through the consistency of neighboring sensors, and the correlation factor is extracted from the association weight matrix of the data gene map.

[0012] Furthermore, the spatial temperature non-uniformity of the abnormal diagnosis and early warning module is obtained by calculating the spatial integral of the temperature field and the average temperature deviation.

[0013] A second aspect of the present invention provides an intelligent monitoring method for cold storage environment based on multi-source data fusion, implemented based on the system, the method comprising the following steps: S1. Multi-source heterogeneous data acquisition and preprocessing: The data acquisition module collects environmental monitoring data, equipment operation data and image monitoring data in the cold storage, and performs Gaussian weighted moving average filtering, local anomaly detection and Z-score standardization on the raw data. S2. Data gene feature extraction and map construction: The statistical domain genes, frequency domain genes and time domain genes of the data stream are extracted through the gene map construction module, the normalized mutual information and maximum information coefficient between gene vectors are calculated, and a multi-layer association weight matrix is ​​constructed. S3. Microclimate fingerprint extraction and pattern recognition: The cold storage space is discretized by the microclimate fingerprint recognition module, and the three-dimensional environmental field is reconstructed by Kriging interpolation. Environmental fingerprint features are extracted and matched with the fingerprint database for recognition. S4. Cargo thermal inertia modeling and temperature calculation: A heat conduction model is established through the thermal inertia compensation module, the temperature distribution equation is solved, the mapping relationship between surface temperature and core temperature is calculated, and temperature compensation parameters are generated. S5. Multi-level adaptive data fusion: The intelligent fusion decision module executes fusion algorithms at the data layer, feature layer, and decision layer respectively, and generates comprehensive environmental situation information by integrating gene similarity, fingerprint matching results, and thermal inertia compensation parameters. S6. Anomaly Pattern Recognition and Risk Assessment: Anomaly feature vectors are constructed through the anomaly diagnosis and early warning module, matched with the pattern library to identify anomaly events, assess the risk level, and generate graded early warning information. S7. Visualization and Quality Assessment of Fusion Results: The visualization module generates a three-dimensional temperature field thermogram and parameter trend curves, and calculates fusion quality assessment indicators to optimize system parameters.

[0014] Furthermore, in step S2, the gene mapping module determines the causal relationship between data sources using the F-statistic of the Granger causality test.

[0015] Furthermore, in step S3, the microclimate fingerprint recognition module uses an anisotropic semivariogram to consider the spatial correlation differences in different directions.

[0016] Furthermore, in step S5, when the evidence conflict coefficient is close to 1, the intelligent fusion decision module introduces a discount factor to correct the basic probability allocation.

[0017] Furthermore, in step S6, the abnormal diagnosis and early warning module determines the equipment fault by the deviation between the amplitude of the fault characteristic frequency in the vibration spectrum and the normal state. Beneficial effects

[0018] 1. Enhanced deep data fusion capabilities: By constructing a data gene map, linear, nonlinear and causal relationships between multi-source data are explored, achieving a technological upgrade from simple numerical fusion to intelligent correlation fusion, which significantly improves data utilization and fusion accuracy.

[0019] 2. Comprehensive environmental perception coverage: By adopting microclimate fingerprint technology, the complete environmental field is reconstructed from sparse sensor data through Kriging interpolation, eliminating monitoring blind spots and realizing all-round three-dimensional monitoring of the cold storage environment.

[0020] 3. Improved accuracy of cargo protection: By introducing a cargo thermal inertia compensation mechanism, taking into account the thermophysical characteristics and response delay of the cargo, the transformation from environmental monitoring to cargo status monitoring has been realized, making the control strategy more aligned with the actual cargo protection needs.

[0021] 4. Improved accuracy of anomaly detection: Multi-dimensional data fusion provides rich feature information, which, combined with pattern recognition and fuzzy logic technology, reduces the false alarm rate and false negative rate, and improves the accuracy of anomaly diagnosis.

[0022] 5. Enhanced system robustness and reliability: Even if some sensors fail, the system can still maintain normal operation through data gene association and environmental fingerprint calculation, which improves the system's fault tolerance and operational reliability. Attached Figure Description

[0023] Figure 1 A schematic diagram of the system architecture described in this invention is shown; Figure 2 A flowchart illustrating the steps of the method described in this invention is shown. Detailed Implementation

[0024] Exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0025] On the one hand, combined with Figure 1This invention provides an intelligent monitoring system for cold storage environments based on multi-source data fusion. Through the synergistic effect of three technical steps—constructing a data genomic map, identifying microclimate fingerprints, and compensating for the thermal inertia of goods—it achieves precise monitoring of the cold storage environment. The system includes a data acquisition module, a genomic map construction module, a microclimate fingerprint identification module, a thermal inertia compensation module, an intelligent fusion decision-making module, an anomaly diagnosis and early warning module, and a visualization display module.

[0026] The data acquisition module is responsible for acquiring environmental monitoring data, equipment operation data, and image monitoring data within the cold storage facility, and preprocessing the raw data. The sensor deployment strategy is determined based on the cold storage space dimensions and monitoring accuracy requirements. For a cold storage facility with dimensions of [length, width, and height]... , , In a cold storage space, the temperature difference between adjacent monitoring points and the spatial distance satisfy the following relationship: in: The temperature difference between adjacent monitoring points; The spatial distance between monitoring points; This is the spatial correlation coefficient, which is related to the insulation performance and airflow organization of the cold storage. It is a power exponent, with a value ranging from 0.5 to 1.5.

[0027] Based on the above relationships, the sensor grid size It should meet the following requirements: in: For sensor grid size; The maximum allowable temperature monitoring error; This is the spatial correlation coefficient; It is a power exponent.

[0028] Number of sensors required for:

[0029] in: Minimum number of sensors; This refers to the length of the cold storage room. Width of the cold storage unit; This refers to the height of the cold storage room. For sensor grid size; It is a rounding function; This represents the redundancy rate, with a value ranging from 0.2 to 0.3.

[0030] When a platinum resistance thermometer is used as the temperature sensor, its resistance value is related to temperature as follows: in: For temperature The resistance value at that time; The resistance value at 0°C; Temperature in Celsius; °C ; °C ;when hour °C ,when hour .

[0031] Sampling interval for data acquisition Determined based on the dynamic characteristics of the signal: in: The sampling interval; The temperature response time constant; The highest frequency component of the signal; This is a function that takes the minimum value.

[0032] The preprocessing of raw data includes three steps: filtering, outlier detection, and standardization. Filtering uses a weighted moving average. in: For the first Filtered data at each time point; For the first The raw data at each moment; For the first The raw data at each moment; For the first The weight coefficients for each position; Length of a single-sided window; weighting coefficient Gaussian distribution is used: in: For the first The weight of each position; This is the position index relative to the center point; The standard deviation of the Gaussian function; It is an exponential function.

[0033] Outlier detection uses the local outlier factor algorithm:

[0034] in: For point based on Local anomalies in the vicinity; For point of Nearest neighbor set; A point in the nearest neighbor set; For point Locally achievable density; For point Locally achievable density; It is the cardinality of the nearest neighbor set.

[0035] Locally achievable density Defined as: in: For point Locally achievable density; For point of Number of nearest neighbors; For point Time The reachable distance is defined as: in: For point Time The reachable distance; For point To its first The distance to the nearest neighbor; For point and points The Euclidean distance between them; This is the function for finding the maximum value.

[0036] Data standardization employs Z-score transformation: in: The value is the standardized value; The original data value; The mean of the data sequence; denoted as the standard deviation of the data sequence.

[0037] The gene mapping module extracts features from the preprocessed data stream and constructs an association map. This module uses wavelet transform for multi-scale analysis. in: These are wavelet coefficients; Scale factor; The translation factor; It is a time-domain signal; For the mother wavelet function; The complex conjugate of the mother wavelet function; It is a time variable.

[0038] In the statistical domain, extract the moment features of the data. The first moment is the mean: in: The mean; This represents the total number of data points. For the first One data point; For data point indices, values ​​range from 1 to... .

[0039] The second-order central moment is the variance: in: For variance; This represents the total number of data points. For the first One data point; This represents the data mean. Index the data points.

[0040] The third-order standard moment is the skewness: in: Skewness; This represents the total number of data points. For the first One data point; The mean; Standard deviation; Index the data points.

[0041] The fourth standard moment is kurtosis: in: For kurtosis; This represents the total number of data points. For the first One data point; The mean; Standard deviation; This is the index for the data points; subtracting 3 is to make the kurtosis of the normal distribution zero.

[0042] Information entropy serves as a measure of data complexity: in: Information entropy; To quantify the total number of intervals; For the data to fall in the first The probability of each interval; For range index; It is a logarithm with base 2.

[0043] In the frequency domain, features are extracted using the Discrete Fourier Transform:

[0044] in: For the first Complex values ​​of each frequency component; For the first One time-domain sampling point; This represents the total number of sampling points; For frequency indexing, values ​​range from 0 to... ; For time-domain indexing, values ​​range from 0 to... ; The imaginary unit; Pi; It is a natural constant.

[0045] The power spectral density is: in: For the first Power spectral density of each frequency component; This represents the total number of sampling points; For complex numbers The modulus; For frequency indexing.

[0046] Extracting the main frequency from the power spectrum : in: Main frequency; For frequency Power spectral density at; Indicates to make Frequency of reaching the maximum value .

[0047] Spectral centroid : in: The centroid of the spectrum; This is the frequency value; For frequency Find the power spectral density at a given point; sum the values ​​across all frequency components.

[0048] In the time domain, an autoregressive moving average model is used: in: for The signal value at that moment; for The signal value at that moment; For the first Autoregressive coefficients; The order of autoregression; For the first Moving average coefficient; The moving average order; for Time-based white noise; for Time-based white noise; For autoregressive term indexing, from 1 to ; For the moving average item index, from 1 to .

[0049] The correlation between different data sources is measured by mutual information: in: For data source and data source Mutual information between them; For data source Gene vectors; For data source Gene vectors; for Gene elements in; for Gene elements in; for and The joint probability distribution of ; for The marginal probability distribution; for The marginal probability distribution; It is the natural logarithm.

[0050] Normalized mutual information is: in: For data source and data source Normalized mutual information between them; The mutual information between the two data sources; For data source The entropy of gene vector information; For data source The entropy of gene vector information.

[0051] Information entropy The calculation formula is: in: Gene vector Information entropy; These are elements in the gene vector; For elements The probability of; It is the natural logarithm.

[0052] The maximum information coefficient is used to measure nonlinear correlation. in: For variables and variables The maximum information coefficient between them; for Number of grid divisions along the axial direction; for Number of grid divisions along the axial direction; This is an upper bound on the number of grid cells. The number of samples; For variables and Mutual information content; for and The minimum value in; It is the natural logarithm.

[0053] Upper bound of grid number Defined as: in: This is an upper bound on the number of grid cells. 0.6 is the sample size; 0.6 is an empirical parameter.

[0054] The constructed multi-layer association weight matrix is ​​as follows: in: This is the total correlation weight matrix; It is a linear correlation matrix; It is a nonlinear correlation matrix; This is a causal relationship matrix; These are the linear correlation weighting coefficients; These are non-linear correlation weighting coefficients; These are the causal relationship weighting coefficients; the three weighting coefficients satisfy the constraints. .

[0055] The microclimate fingerprint recognition module discretizes the cold storage space into micro-environmental units and reconstructs the complete environmental field through spatial interpolation. Unit size... Determined dynamically based on temperature gradient: in: For microenvironment unit size; The allowable temperature difference threshold; This represents the magnitude of the maximum temperature gradient. Maximum unit size limit; This is a function that takes the minimum value.

[0056] Temperature gradient The calculation formula is: in: This is the temperature gradient vector; For temperature at Partial derivatives in direction; For temperature at Partial derivatives in direction; For temperature at Partial derivatives in direction.

[0057] For areas not covered by the sensor, an improved Kriging interpolation method is used. First, an anisotropic semivariogram is established: in: For the separation vector The corresponding semi-mutated value; This represents the nugget value, indicating variation at the microscale. The sill value represents the overall level of variation. A three-dimensional separation vector; , , They are respectively , , The separation distance in direction; , , These are the range parameters for the three directions; It is an exponential function.

[0058] Kriging weights are obtained by solving the following system of linear equations: in: For the first The monitoring point and the first The semi-variance value among the monitoring points; For the first The semi-variance between each monitoring point and the point to be estimated; For the first Kriging weights for each monitoring point; This represents the total number of monitoring points. It is a Lagrange multiplier; , This is the index for the monitoring point, with values ​​ranging from 1 to... .

[0059] semi-mutated value The calculation formula is: in: For point and points The semi-variance values ​​between; For the first Spatial coordinate vectors of each monitoring point; For the first Spatial coordinate vectors of each monitoring point; It is a semi-variogram.

[0060] The interpolation estimate is: in: For position The estimated value; For the first Observations from each monitoring point; For the first The weight of each monitoring point; This represents the total number of monitoring points. For monitoring point index.

[0061] Based on the reconstructed environmental field, static fingerprint features of each micro-environment unit are extracted: in: This is a static fingerprint feature vector; The average temperature of the unit; This represents the magnitude of the temperature gradient. For the Laplace operator of temperature; It is the airflow direction angle; This represents the magnitude of the relative humidity gradient.

[0062] Temperature Laplace operator Defined as: in: For temperature Laplace operator; For temperature at Second-order partial derivatives in the direction; For temperature at Second-order partial derivatives in the direction; For temperature at Second-order partial derivatives in the direction.

[0063] Dynamic fingerprint features include: in: This is a dynamic fingerprint feature vector; This is the first-order partial derivative of temperature with respect to time; This is the second partial derivative of temperature with respect to time; It is the airflow velocity vector; Let be the divergence of the velocity field; Let be the curl of the velocity field.

[0064] velocity field divergence The calculation formula is: in: Let the velocity field divergence be denoted as . For speed in Component of direction; For speed in Component of direction; For speed in Component of direction; , , These are the partial derivatives of each velocity component with respect to the corresponding coordinate.

[0065] Velocity field curl The calculation formula is: in: Let be the curl vector of the velocity field; each component is the curl component in the corresponding direction.

[0066] Fingerprint matching uses cosine similarity: in: The similarity between two fingerprint vectors; This is the first fingerprint vector; This is the second fingerprint vector; It is the dot product of two vectors; For vectors The Euclidean norm; For vectors The Euclidean norm.

[0067] The thermal inertia compensation module considers the impact of the cargo's thermophysical properties on temperature monitoring. The internal temperature evolution of the cargo follows the heat conduction equation: in: For cargo density; Specific heat capacity of the goods; For temperature field; For time; Thermal conductivity; For the Laplace operator; This is an internal heat source item.

[0068] For composite material cargo, the equivalent thermophysical parameters are calculated using volume weighting: in: Equivalent density; This is the equivalent specific heat capacity; It is the equivalent thermal conductivity; For the first Volume fraction of the material; For the first The density of the material; For the first The specific heat capacity of the material; For the first The thermal conductivity of this material; Number of material types; For material index.

[0069] For irregularly stacked goods, the effect of porosity should be considered: in: The equivalent thermal conductivity of the stack; The thermal conductivity of solid cargo; The thermal conductivity of air; Porosity.

[0070] The temperature response of the cargo is described using a lumped parameter model: in: For cargo temperature; For time; The convective heat transfer coefficient; The surface area of ​​the goods; For the quality of goods; Specific heat capacity of the goods; The ambient temperature; The internal heat generation rate of the cargo.

[0071] The convective heat transfer coefficient is determined by the Nusselt number correlation: in: For Nusselt numbers; It is the Reynolds number; is the Prandtl number; the exponents 1 / 2 and 1 / 3 are empirical coefficients.

[0072] The definitions of Nusselt numbers, Reynolds numbers, and Prandtl numbers are as follows: in: The convective heat transfer coefficient; The characteristic length; The thermal conductivity of air; air density; The airflow velocity; Aerodynamic viscosity; This refers to the specific heat capacity of air at constant pressure.

[0073] Establish a dynamic mapping relationship between surface temperature and core temperature: in: for The core temperature of the cargo at any given moment; for Surface temperature at any given time; The thermal response time constant; The surface temperature change rate; The attenuation coefficient; For integration variables; is a natural constant; the integral term represents the cumulative effect of historical temperatures.

[0074] The intelligent fusion decision-making module receives information from the gene mapping module, the microclimate fingerprinting module, and the thermal inertia compensation module, and performs multi-level data fusion. In the data layer fusion, the reliability of each sensor is comprehensively evaluated through three factors: in: For the first The fusion weights of the individual sensors; For the first The reputation factor of each sensor; For the first The reliability factor of each sensor; For the first The correlation factors of each sensor; This represents the total number of sensors of the same type. For sensor indexing, from 1 to .

[0075] Reputation Factor Calculation based on historical prediction error: in: For the first The reputation factor of each sensor; To assess the length of the time window; For time indexing; For the first Each sensor at time The measured value; For the first Each sensor at time The predicted value; It is an exponential function.

[0076] Credibility factor Determined by the consistency of neighboring sensors:

[0077] in: For the first The reliability factor of each sensor; Number of nearest neighbors; For sensors of Nearest neighbor set; Nearest neighbor index; For the first The current measurement value of each sensor; For the first Measurements from nearby sensors; It is the maximum of the absolute values ​​of the two.

[0078] Related factors Association weights derived from gene maps: in: For the first The correlation factors of each sensor; This represents the total number of sensors; Sensors in the gene map and sensors The correlation weight between them; For sensor indexing.

[0079] The data layer fusion result is as follows: in: The merged data values; For the first The fusion weights of the individual sensors; For the first Measurement values ​​from each sensor; This represents the total number of sensors; For sensor indexing.

[0080] In feature layer fusion, feature vectors from different sources are concatenated: in: The concatenated feature vector; Gene feature vector; This is an environmental fingerprint vector; This is the thermal inertia eigenvector.

[0081] Principal component analysis was used for dimensionality reduction. in: These are the eigenvectors after dimensionality reduction; Principal component matrix; This is the original feature vector; The feature mean vector; superscript This indicates the matrix transpose.

[0082] The principal component matrix is ​​obtained through eigenvalue decomposition: in: The characteristic covariance matrix; The eigenvector matrix; It is an eigenvalue diagonal matrix; for The transpose of .

[0083] Before choosing The eigenvectors corresponding to the largest eigenvalues ​​satisfy the following cumulative variance contribution rate: in: For the first One eigenvalue; The number of principal components selected; The original feature dimension; For feature value index.

[0084] In decision-level fusion, the Dempster-Shafer evidence theory is adopted. The identification framework is defined as follows: in: For identification framework; Indicates a normal state; express An abnormal state.

[0085] The basic probability assignment function satisfies: in: For the first A subset of evidence sources The basic probability distribution; for any subset of; It is an empty set.

[0086] Dempster's combination rules are: in: For the combined subset The probability allocation; For the first source pair subset of evidence The probability allocation; For the second source of evidence pair subset The probability allocation; and for A subset of; This represents the conflict coefficient.

[0087] Conflict coefficient Defined as: in: The conflict coefficient; and To satisfy the condition of subset pairs with an empty intersection.

[0088] When conflicting evidence exists, a discount factor is introduced: in: Assigning probabilities after discounting; Assign the original probabilities; For the first Discount factor for each source of evidence; It is uniformly distributed.

[0089] Uniform distribution is defined as: in: It is a uniform distribution function; To identify the cardinality of the frame; express The size of the power set.

[0090] The anomaly diagnosis and early warning module identifies anomalies based on the fused comprehensive situational information. Temperature anomaly patterns are described by feature vectors: in: This represents the temperature anomaly pattern vector. The magnitude of the temperature deviation from the set value; The rate of temperature change; This represents the standard deviation of temperature fluctuation. This refers to spatial temperature non-uniformity. The duration of the abnormality.

[0091] Spatial temperature non-uniformity Defined as: in: This refers to spatial temperature non-uniformity. For cold storage volume; For position Temperature at that location; The average temperature in space; It is a volumetric infinitesimal element.

[0092] Equipment faults are identified through vibration spectrum characteristics. The characteristic frequencies of faults in rotating equipment are: in: The fault characteristic frequency; For harmonic order; For rotational speed and frequency; Slip rate; It represents two possibilities: positive and negative.

[0093] The fault determination criteria are: in: The amplitude at the fault characteristic frequency; This is the average amplitude of this frequency under normal conditions; This represents the standard deviation of the frequency amplitude under normal conditions. is the confidence coefficient, with a value ranging from 2 to 3.

[0094] The cumulative damage model is used for cargo quality risk assessment. in: For cumulative risk value; The current moment; For integration variables; The time weighting function; This is the damage function due to temperature deviation. for Temperature at any moment; The optimal storage temperature.

[0095] The time weighting function uses exponential decay: in: for The weight of each moment; The current moment; The decay time constant; It is a natural constant.

[0096] For perishable goods, quality degradation follows the Arrhenius equation: in: For quality indicators; For time; For frequency factors; It is the activation energy; The ideal gas constant is 8.314 J / (mol·K); Absolute temperature; It is an exponential function.

[0097] The warning level is determined using fuzzy logic rules. The membership function for temperature deviation is defined as follows: in: Temperature deviation Membership degree; , , , This is the turning point of the membership function.

[0098] The overall warning level is: in: The overall warning level is [level]. The total number of rules; For rule indexing; For the first The membership degree of a rule; For the first The warning level value corresponding to each rule; The input feature vector; This is the function for finding the maximum value.

[0099] The visualization module converts monitoring data and analysis results into a visual format. The three-dimensional temperature field is rendered using volumetric rendering technology. in: For rays The color value of the direction; The depth of the ray-traversed body data; These are the position parameters on the ray; For position Opacity at the location; For position The color of the place; For integration variables; It is an exponential function.

[0100] The temperature-to-color mapping uses a piecewise linear transfer function: in: For temperature The corresponding color value; The RGB value is for blue. The RGB value is green. The RGB value is for red. This is the lowest temperature threshold. The optimal temperature; This represents the highest temperature threshold.

[0101] Based on the above system architecture, combined with Figure 2 The present invention also provides an intelligent monitoring method for cold storage environment based on multi-source data fusion, the method comprising the following steps: S1. Multi-source heterogeneous data acquisition and preprocessing The data acquisition module collects environmental monitoring data, equipment operation data, and image monitoring data from the cold storage facility. The raw data is then processed using moving average filtering, outlier detection, and standardization to generate a data stream in a unified format, providing a high-quality data foundation for subsequent processing.

[0102] S2, Data Gene Feature Extraction and Map Construction Using the gene map construction module, statistical domain, frequency domain, and time domain features are extracted from the preprocessed data stream as data genes. The mutual information and maximum information coefficient between genes from different data sources are calculated, and a multi-layer association weight matrix reflecting the intrinsic relationship of the data is constructed to form a complete data gene map.

[0103] S3, Microclimate Fingerprint Extraction and Pattern Recognition The cold storage space is discretized into micro-environmental units using a microclimate fingerprinting module. A complete three-dimensional environmental field is reconstructed using Kriging interpolation. The mean temperature, temperature gradient modulus, temperature Laplace operator, airflow direction angle, relative humidity gradient modulus, temperature-time rate of change, second-order temperature-time rate of change, velocity field divergence, and velocity field curl of each unit are extracted as environmental fingerprint features. These features are then matched with standard and abnormal fingerprint databases to identify the current environmental operating mode.

[0104] S4. Cargo Thermal Inertia Modeling and Temperature Calculation Using a thermal inertia compensation module, a heat conduction model is established based on the cargo type and stacking method. By solving the heat conduction equation and the lumped parameter model, the dynamic mapping relationship between the internal temperature and the surface temperature of the cargo is calculated, and temperature compensation parameters that take into account the thermal inertia of the cargo are generated.

[0105] S5, Multi-level Adaptive Data Fusion The intelligent fusion decision module is adopted. Based on data gene similarity, environmental fingerprint matching results and thermal inertia compensation parameters, it integrates similar sensor data through credibility weighting at the data layer, integrates multi-source features through principal component analysis for dimensionality reduction at the feature layer, and integrates the judgment results of multiple subsystems through DS evidence theory at the decision layer to generate highly credible comprehensive environmental situation information.

[0106] S6. Anomaly Pattern Recognition and Risk Assessment Using the anomaly diagnosis and early warning module, based on the integrated situational information, the system identifies temperature anomaly events, equipment failure events, and cargo quality risk events by matching them with temperature anomaly pattern libraries, equipment failure pattern libraries, and cargo risk pattern libraries. Fuzzy logic rules are used to determine the early warning level, and graded early warning information and control suggestions are generated.

[0107] S7. Visualization and Quality Assessment of Fusion Results The visualization module uses volume rendering technology to generate a three-dimensional temperature field thermogram, plotting trend curves of key parameters and a list of abnormal events. Simultaneously, it calculates the fusion gain rate, information entropy reduction rate, and system robustness index to evaluate the fusion effect. Based on the evaluation results, the fusion parameters are adjusted to achieve continuous system optimization.

[0108] In summary, this invention provides an intelligent monitoring system and method for cold storage environment based on multi-source data fusion. By constructing a data gene map, identifying microclimate fingerprints, and compensating for the thermal inertia of goods, it achieves accurate monitoring and intelligent control of the cold storage environment, effectively solving the technical problems of low data fusion, incomplete environmental perception, and neglect of cargo characteristics in the prior art.

[0109] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A cold storage environment intelligent monitoring system based on multi-source data fusion, characterized in that, include: The data acquisition module is used to collect environmental monitoring data, equipment operation data, and image monitoring data within the cold storage facility and to perform preprocessing. The gene map construction module is used to extract statistical domain, frequency domain and time domain features from the preprocessed data stream as data genes, and construct a multi-layer association weight matrix by calculating the mutual information between genes from different data sources to form a data gene map. The microclimate fingerprinting module is used to discretize the cold storage space into micro-environment units, reconstruct the three-dimensional environmental field using spatial interpolation, and extract the temperature distribution, gradient field, and dynamic change characteristics of each micro-environment unit as environmental fingerprints. The thermal inertia compensation module is used to establish a heat conduction model based on the cargo type, calculate the dynamic mapping relationship between the internal temperature and surface temperature of the cargo, and generate temperature compensation parameters that take into account the thermal response delay of the cargo. The intelligent fusion decision module is used to generate comprehensive situational information of the cold storage environment by executing fusion algorithms at the data layer, feature layer, and decision layer based on data gene similarity, environmental fingerprint matching results, and thermal inertia compensation parameters.

2. The system according to claim 1, characterized in that, The statistical domain features extracted by the gene mapping module include mean, variance, skewness, and kurtosis; the frequency domain features include dominant frequency and power spectral density; and the time domain features include autoregressive model coefficients.

3. The system according to claim 1, characterized in that, The microclimate fingerprint recognition module uses the Kriging interpolation method to obtain interpolation weights by solving a system of linear equations determined by the semi-variogram function, and then reconstructs the complete three-dimensional temperature field.

4. The system according to claim 1, characterized in that, The thermal inertia compensation module establishes a mapping relationship between surface temperature and core temperature, including thermal response time constant and the cumulative influence of historical temperature, by solving the partial differential equation of heat conduction and the lumped parameter model.

5. The system according to claim 1, characterized in that, The intelligent fusion decision-making module calculates the fusion weight at the data layer using reputation factor, credibility factor, and correlation factor. The reputation factor is based on historical prediction error, the credibility factor is based on the consistency of neighboring sensors, and the correlation factor is based on the correlation of gene maps.

6. The system according to claim 1, characterized in that, It also includes an anomaly diagnosis and early warning module, which generates graded early warning information based on the fused comprehensive situational information by constructing temperature anomaly pattern vectors, identifying equipment fault characteristic frequencies, and assessing cumulative cargo damage.

7. A method for intelligent monitoring of cold storage environment based on multi-source data fusion, characterized in that, Includes the following steps: S1. Collect environmental monitoring data, equipment operation data, and image monitoring data from the cold storage, and perform filtering, outlier detection, and standardized preprocessing. S2. Extract statistical domain, frequency domain, and time domain features from the preprocessed data stream as data genes, calculate the mutual information between genes from different data sources, and construct a gene map that reflects the intrinsic relationship of the data. S3. Discretize the cold storage space into micro-environmental units, reconstruct the three-dimensional environmental field through spatial interpolation, extract environmental fingerprint features and match them with the fingerprint database; S4. Establish a heat conduction model based on the cargo type, calculate the internal temperature distribution of the cargo, and generate temperature compensation parameters; S5. By combining the gene similarity of the integrated data, the environmental fingerprint matching results, and the thermal inertia compensation parameters, multi-level data fusion is performed to generate comprehensive environmental situation information.

8. The method according to claim 7, characterized in that, In step S2, a multi-layered correlation weight matrix containing linear correlation, nonlinear correlation, and causal relationship is constructed by calculating normalized mutual information and maximum information coefficient.

9. The method according to claim 7, characterized in that, In step S3, the environmental fingerprint features include the mean temperature, the magnitude of the temperature gradient, the Laplacian operator of temperature, the rate of change of temperature over time, the velocity field divergence, and the velocity field curl.

10. The method according to claim 7, characterized in that, In step S5, evidence theory is used to integrate the decision-making level. When the evidence conflict coefficient is close to 1, a discount factor is introduced to correct the basic probability allocation.

Images