Remote platform monitoring system and method for refrigeration house refrigerating unit
By combining multi-source data acquisition with granular computing theory and LSTM model, the problems of delayed prediction cycle and insufficient accuracy of cold storage refrigeration unit status are solved, realizing accurate status monitoring of cold storage refrigeration units and improving the intelligence and reliability of operation and maintenance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HANGZHOU DUOXIE INFORMATION TECH CO LTD
- Filing Date
- 2026-01-07
- Publication Date
- 2026-04-21
AI Technical Summary
Existing cold storage refrigeration units have problems with delayed prediction cycles and insufficient accuracy in short prediction cycles, which may cause medicines to become ineffective at abnormal temperatures and affect their quality.
By combining multi-source data acquisition with granular computing theory and the LSTM model, an ultra-short-term state prediction model is constructed through data preprocessing, granular feature processing, cross-granularity correlation analysis, and state prediction, thereby achieving accurate state monitoring of cold storage refrigeration units.
It enables accurate status prediction of cold storage refrigeration units, improves the level of intelligent operation and maintenance, reduces the probability of downtime and operation and maintenance costs, and ensures the safety and controllability of the cold storage environment.
Smart Images
Figure CN121898097A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of cold storage refrigeration unit monitoring technology, specifically a remote platform monitoring system and method for cold storage refrigeration units. Background Technology
[0002] Medicines have high requirements when they are kept warm, but the refrigeration units in cold storage for medicines may not meet the refrigeration requirements during use, thus failing to meet the refrigeration requirements of medicines. The existing technical problems in the prediction of refrigeration status of cold storage refrigeration units are: (1) the status prediction cycle is too late; (2) the accuracy of the status in the short prediction cycle is not enough. The existing technical problems often lead to the loss or reduction of drug efficacy at abnormal temperatures. In order to solve this technical problem, there is an urgent need for a remote monitoring solution for cold storage refrigeration units that can achieve refined processing of multi-source data, deep correlation analysis of multi-granular features, and accurate state prediction by combining intelligent algorithms. This solution can make up for the deficiencies of existing technologies and improve the intelligence and reliability of cold storage refrigeration unit operation and maintenance.
[0003] Therefore, a remote platform monitoring system and method for cold storage refrigeration units are provided. Summary of the Invention
[0004] To address the aforementioned technical problems, the present invention aims to provide a remote platform monitoring system and method for cold storage refrigeration units.
[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for remote platform monitoring of a cold storage refrigeration unit, the method comprising: Collect multi-source operating status data of the cold storage refrigeration unit and cold storage environment data, and perform data preprocessing on the multi-source operating status data and cold storage environment data to obtain corresponding standardized operating status data and standardized cold storage environment data. Based on granular computing theory, granular feature processing is performed on standardized operating status data and standardized cold storage environment data to obtain corresponding multi-granularity monitoring feature datasets. Perform cross-granularity correlation analysis on the multi-granularity monitoring feature dataset to obtain the corresponding global fusion feature difference; Based on LSTM and according to the global fusion feature difference, an ultra-short-term state prediction model for cold storage refrigeration units is obtained. Based on the ultra-short-term state prediction model, the operating status dataset of each refrigeration unit within an ultra-short monitoring period is obtained.
[0006] Furthermore, based on granular computing theory, the process of granular feature processing for standardized operational status data and standardized cold storage environment data includes: The standardized operating status data and standardized cold storage environment data are divided into feature granules to obtain the corresponding feature granule levels; Each feature particle is divided into time dimensions using a fixed time window. The length of the time window is configurable. Based on the time window length, the time particle hierarchy corresponding to the time window length is obtained. Based on the time granularity level and the feature granularity level, granular feature processing is performed on standardized operating status data and standardized cold storage environment data to obtain the corresponding multi-granularity monitoring feature dataset.
[0007] Furthermore, based on the time granularity level and the feature granularity level, the process of granular feature processing for standardized operating status data and standardized cold storage environment data includes: Determine the time granularity level and the feature granularity level, using the feature granularity level-time granularity level as a unique identifier; group the standardized operating status data and standardized cold storage environment data according to the unique identifier. Based on the unique identifier, statistical feature processing, trend feature processing, and fluctuation feature processing are performed on each group of data to obtain the statistical feature set, trend feature set, and fluctuation feature set. The statistical feature set includes the mean, variance, standard deviation, peak value, trough value, and range of the corresponding data; the trend feature set includes the linear fit slope and trend correlation coefficient of the corresponding data; and the fluctuation feature set includes the coefficient of variation, fluctuation frequency, and fluctuation amplitude of the corresponding data.
[0008] Furthermore, the process of performing trend feature processing on each group of data includes: During the acquisition period, the starting time of the time window of the time particle in the corresponding time particle level is set to 0, and the time window length is incremented according to the corresponding time particle level to form time series data. Linear regression analysis of time series data is performed using the least squares method to obtain the linear fit slope; the corresponding trend correlation coefficient is obtained based on the Pearson correlation coefficient.
[0009] Furthermore, the process of processing the fluctuation characteristics of each group of data includes: Calculate the coefficient of variation of the corresponding data based on the standard deviation and mean obtained from statistical feature processing; Traverse the time-series data within the information granules, compare the magnitude of two adjacent data points with the mean of the corresponding data points, and count the number of effective crossings; obtain the fluctuation frequency based on the number of effective crossings and the corresponding time window length; obtain the fluctuation amplitude based on the mean of the corresponding data and the data at the actual collection time point.
[0010] Furthermore, the process of performing cross-granularity correlation analysis on the multi-granularity monitoring feature dataset includes: The multi-granularity monitoring feature dataset is divided into fine-grained feature vectors, medium-grained feature vectors, and coarse-grained feature vectors. Calculate the first cross-granularity correlation coefficient based on the fine-grained feature vector and the medium-grained feature vector; calculate the second cross-granularity correlation coefficient based on the fine-grained feature vector and the coarse-grained feature vector. Based on the first cross-granularity correlation coefficient and the second cross-granularity correlation coefficient, the global fusion feature difference for the corresponding acquisition period is obtained; Preset global fusion feature difference threshold; If the global fusion feature difference is less than or equal to the global fusion feature difference threshold, it indicates that the chiller unit is operating normally during the current acquisition period. If the difference in global fusion features is greater than the threshold for global fusion feature difference, it indicates that the operating status of the refrigeration unit in the current acquisition period is abnormal. The multi-source operating status data and cold storage environment data of the corresponding acquisition period are then obtained. Based on the multi-source operating status data and cold storage environment data of the corresponding acquisition period, the operating status of the corresponding equipment is specifically analyzed. If there is an abnormality, it will be marked and a corresponding set of abnormal equipment statuses will be formed; otherwise, the corresponding equipment operating status is normal.
[0011] Furthermore, the process of constructing an ultra-short-term state prediction model includes: Obtain the global fusion feature difference and the global fusion feature difference threshold for several historical acquisition periods; The global fusion feature differences and global fusion feature difference thresholds from several historical acquisition periods are grouped and labeled, denoted as follows: It is a natural number; Pick The global fusion feature difference and the global fusion feature difference threshold of the historical acquisition period are used as the sample set, and Less than The natural number is used as the test set, and the global fusion feature difference and global fusion feature difference threshold of the remaining historical collection periods are used as the test set. The training sample set is formed based on the sample set and the test set.
[0012] A standard state prediction model is constructed based on LSTM. The training sample set is input into the standard state prediction model for training, and the trained standard state prediction model is denoted as the ultra-short-term state prediction model.
[0013] A second aspect of the present invention also provides a remote platform monitoring system for cold storage refrigeration units, comprising: a data acquisition module, a data preprocessing module, a particle feature processing module, a cross-particle size analysis module, and a state prediction module; The data acquisition module is used to collect multi-source operating status data of the cold storage refrigeration unit and cold storage environmental data; The data preprocessing module is used to preprocess multi-source operating status data and cold storage environment data to obtain corresponding standardized operating status data and standardized cold storage environment data. The granular feature processing module, based on granular computing theory, performs granular feature processing on standardized operating status data and standardized cold storage environment data to obtain corresponding multi-granularity monitoring feature datasets. The cross-granularity analysis module is used to perform cross-granularity correlation analysis on multi-granularity monitoring feature datasets to obtain the corresponding global fusion feature differences; The state prediction module, based on LSTM, obtains an ultra-short-term state prediction model for the cold storage refrigeration units according to the global fusion feature difference, and obtains the operating state dataset of each refrigeration unit within an ultra-short monitoring period based on the ultra-short-term state prediction model.
[0014] Compared with existing technologies, the beneficial effects of this invention are as follows: By collecting multi-source operational status data and cold storage environment data and performing standardized preprocessing, comprehensive coverage and unified standardization of monitoring data are achieved, laying a high-quality data foundation for subsequent data processing and analysis, and effectively solving the problems of single data collection dimensions and insufficient standardization in existing technologies. Based on granular computing theory, standardized data is hierarchically divided into feature granules and time granules, and multi-dimensional features are extracted to form a multi-granularity monitoring feature dataset, realizing refined hierarchical processing of data, fully mining the feature information of data at different levels, and improving the representation ability of data features; and by constructing a global fusion feature difference through cross-granularity correlation analysis, deep fusion and correlation verification of feature information at different time scales are achieved, improving the accuracy of anomaly identification and avoiding the limitations of traditional fixed threshold judgment. Based on LSTM, an ultra-short-term state prediction model is constructed, and the global fusion feature difference is used for model training and prediction, which can predict the operating status of refrigeration units in advance, providing accurate early warning information for operation and maintenance personnel, helping to achieve preventive maintenance, reducing the probability of failure downtime and operation and maintenance costs, and ensuring the stable operation of cold storage refrigeration units and the safe and controllable cold storage environment. Attached Figure Description
[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.
[0016] Figure 1 This is a schematic diagram illustrating the steps of a remote platform monitoring method for cold storage refrigeration units.
[0017] Figure 2This is a schematic diagram of a remote platform monitoring system for cold storage refrigeration units. Detailed Implementation
[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.
[0019] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0020] Example 1: like Figure 1 As shown, a method for remote monitoring a cold storage refrigeration unit includes the following steps: Step S1: Collect multi-source operating status data of the cold storage refrigeration unit and cold storage environment data, and perform data preprocessing on the multi-source operating status data and cold storage environment data to obtain corresponding standardized operating status data and standardized cold storage environment data. In the actual data acquisition process, different types of data can be collected through different sensors. For example, a platinum resistance sensor can be placed close to the measured point on the compressor exhaust pipe to ensure good heat conduction and then output a resistance signal. This signal is then converted into a standard voltage signal or a standard current signal by a temperature transmitter and transmitted to a PLC (Programmable Logic Controller). The PLC is a digital computing and operating electronic system used to realize numerical reading and recording.
[0021] Using corresponding data acquisition sensors, various data acquisition points are set up to collect real-time data such as compressor discharge temperature, real-time motor current, real-time power, condenser pressure, refrigerant flow rate, and evaporator pressure of the cold storage refrigeration unit, as well as real-time temperature and humidity data of the cold storage. These various data points related to the cold storage refrigeration unit are collectively recorded as multi-source operating status data; and various data points related to the cold storage environment are collectively recorded as cold storage environment data. In this embodiment, multi-source operating status data includes, but is not limited to, real-time compressor discharge temperature, real-time motor current, real-time power, condenser pressure, refrigerant flow rate, and evaporator pressure; cold storage environment data includes, but is not limited to, real-time temperature and humidity data.
[0022] It should be noted that the acquisition frequencies of the corresponding types of data acquisition sensors need to be synchronized to ensure data consistency during data analysis, thereby reducing inaccurate data processing results caused by data deviation. In addition, an acquisition cycle should be set, with each acquisition cycle containing several data acquisition moments, to facilitate the analysis of data characteristics within the corresponding acquisition cycle.
[0023] In practical applications, data acquisition sensors are affected by factors such as actual operating conditions, sensor damage, or data transmission, resulting in abnormal data in the data collected by the corresponding sensors. This abnormal data can affect subsequent data processing or analysis. Therefore, it is necessary to preprocess the collected multi-source operating status data and cold storage environment data to remove abnormal data from the corresponding collection period.
[0024] In the actual data preprocessing process, multi-source operational status data and cold storage environment data corresponding to the collection period are read from the backend system, based on... The principle is to remove any abnormal data that may exist in the multi-source operating status data and cold storage environment data of the corresponding collection period, so as to obtain standardized operating status data and standardized cold storage environment data of the corresponding collection period. The specific data preprocessing will not be described in detail.
[0025] Step S2: Based on granular computing theory, perform granular feature processing on standardized operating status data and standardized cold storage environment data to obtain the corresponding multi-granularity monitoring feature dataset; In practical applications, based on granular computing theory, the steps for granular feature processing of standardized operating status data and standardized cold storage environment data are as follows: A1: Divide the standardized operating status data and standardized cold storage environment data into feature granules to obtain the corresponding feature granule levels; It should be noted that the feature granule hierarchy is divided into compressor operation feature granules, condenser feature granules, evaporator feature granules, and cold storage environment feature granules according to data type. Data within the same feature granule satisfies the attribute similarity criterion. The compressor operation feature granules include, but are not limited to, exhaust temperature features, current features, and power features; the condenser feature granules include, but are not limited to, real-time condenser pressure features and real-time refrigerant flow features; the evaporator feature granules include, but are not limited to, evaporation pressure features; and the cold storage environment feature granules include, but are not limited to, temperature features and humidity features.
[0026] In this embodiment, the data within the same feature granule satisfying the attribute similarity criterion can be understood as follows: after dividing the standardized operating status data and standardized cold storage environment data into feature granules, all data within the same feature granule have similar attribute characteristics, physical meanings, or monitoring targets. They aggregate around the same monitoring object or the same type of monitoring dimension, thereby achieving data classification, merging, and dimensionality reduction. For example, the compressor operation feature granule includes exhaust temperature characteristics, current characteristics, and power characteristics. The common attribute of these data is that they directly reflect the operating conditions of the compressor. The exhaust temperature characteristic is related to the compressor's compression efficiency, while the current and power characteristics are related to the compressor's power input. These three attributes are highly correlated, therefore they are grouped into the same feature granule.
[0027] A2: Divide each feature particle into time dimensions using a fixed time window. The length of the time window is configurable. Based on the length of the time window, obtain the time particle hierarchy corresponding to the length of the time window. In this embodiment, the time granule level includes fine-grained time granules, medium-grained time granules, and coarse-grained time granules, and the corresponding time window lengths can be set to 5s, 15s, 30s, or according to actual needs.
[0028] A3: Based on the time granularity level and the feature granularity level, perform granular feature processing on the standardized operating status data and standardized cold storage environment data to obtain the corresponding multi-granularity monitoring feature dataset.
[0029] In step A3, the steps for granular feature processing of standardized operating status data and standardized cold storage environment data based on time granularity and feature granularity are as follows: A31: Determine the time granularity level and the feature granularity level, using the feature granularity level-time granularity level as a unique identifier; group the standardized operating status data and standardized cold storage environment data according to the unique identifier; It's important to note that the grouping process is not a superposition or overlap of data collection periods. For example, if the time granularity is determined by time window lengths of 5s, 15s, and 30s, then the corresponding data collection times for the collection periods are 5s, 15s, and 30s, respectively. The length of the collection period is fixed; the values are simply standardized operational status data and standardized cold storage environment data collected at the corresponding time granularity. By analyzing the data characteristics at different time granularities, we can further analyze changes in data characteristics, meeting the data support requirements for short prediction cycle accuracy.
[0030] A32: Based on the unique identifier, perform statistical feature processing, trend feature processing, and fluctuation feature processing on the data of each group to obtain the statistical feature set, trend feature set, and fluctuation feature set; In this embodiment, statistical feature processing includes calculating the mean, variance, standard deviation, peak value, trough value, and range of the data; the specific calculation processes for the mean, variance, standard deviation, peak value, trough value, and range are not detailed here. The statistical feature set includes the mean, variance, standard deviation, peak value, trough value, and range of the corresponding data; the trend feature set includes the linear fitting slope and trend correlation coefficient of the corresponding data; the fluctuation feature set includes the coefficient of variation, fluctuation frequency, and fluctuation amplitude of the corresponding data. The statistical feature set, trend feature set, and fluctuation feature set represent the corresponding multi-granularity monitoring feature dataset.
[0031] In step A32, the trend feature processing steps are as follows: A321: During the acquisition cycle, the starting time of the time window corresponding to the time grain in the time grain level of the corresponding feature grain level is set to 0, and incremented according to the corresponding time window length. The obtained timestamp sequence is recorded as follows. The data corresponding to the timestamp sequence is denoted as Composition of time series data ; A322: Time-series data based on least squares method Perform linear regression analysis to obtain the linear regression equation: ;in, This represents data of the corresponding type; This represents the timestamp within the corresponding time window length for the corresponding type. Indicates the slope of the linear fit; This represents the intercept.
[0032] A323: Based on timestamp sequence and the corresponding timestamp sequence data Calculate the corresponding time average. and data mean Then, the slope of the linear fit is solved. ; It should be noted that the change in monitored data per unit time is... This indicates that the data increases over time. This indicates that the data decreases over time. Indicates no clear trend. Intercept It is used to construct a complete linear regression equation and verify the fit effect.
[0033] A324: The calculation process based on the Pearson correlation coefficient to obtain the trend correlation coefficient. .
[0034] It should be noted that the trend correlation coefficient Determining the relationship requires setting corresponding threshold values based on the operating characteristics of the cold storage refrigeration unit. In practical applications, for example, the compressor's exhaust temperature can be represented as shown in the table below: In step A32, the steps for processing the fluctuation characteristics are as follows: A325: Calculate the coefficient of variation of the corresponding data based on the standard deviation and mean obtained from statistical feature processing. ; It should be noted that the coefficient of variation is a dimensionless index of dispersion, used to eliminate the influence of data dimensions and mean size. It is more suitable than the standard deviation for comparing the degree of fluctuation of different types and magnitudes of data.
[0035] It should be noted that the coefficient of variation Determining the relationship also requires setting corresponding threshold values based on the operating characteristics of the cold storage refrigeration unit. In practical applications, for example, the compressor's exhaust temperature can be represented as shown in the table below: A326: Traversing the temporal data within information granules The effective crossing count is calculated by comparing the magnitudes of two adjacent data points with their corresponding mean values. Based on the effective crossing count and the corresponding time window length, the fluctuation frequency is obtained. ;in, Indicates the number of valid traversals; Indicates the length of the time window.
[0036] In practical application, if ; indicates the first To the If a data point is successfully traversed, the count of successful traversals is incremented by one. or This indicates that the data point falls on the mean line and is not counted as a valid crossing point.
[0037] A327: Obtain the fluctuation range based on the mean of the corresponding data and the data at the actual collection time point. ;in, This indicates the maximum deviation of the corresponding data from the mean; This represents the minimum deviation of the corresponding data from the mean.
[0038] For example, assuming the time granularity is determined by time window lengths of 5s, 15s, and 30s, then after performing statistical feature processing, trend feature processing, and fluctuation feature processing on the data in the corresponding collection period, there will be three sets of data: mean, variance, standard deviation, peak value, trough value, and range for statistical feature processing under time window lengths of 5s, 15s, and 30s; linear fit slope and trend correlation coefficient for trend feature processing; and coefficient of variation, fluctuation frequency, and fluctuation amplitude for fluctuation feature processing. Since there are several standardized operating status data and standardized cold storage environment data collected, the compressor operation feature granules, condenser feature granules, evaporator feature granules, and cold storage environment feature granules in the actual feature granularity level each contain a corresponding number of data features.
[0039] Step S3: Perform cross-granularity correlation analysis on the multi-granularity monitoring feature dataset to obtain the corresponding global fusion feature difference; In practical applications, based on the time granularity level, the multi-granularity monitoring feature dataset is divided into fine-grained feature vectors, medium-grained feature vectors, and coarse-grained feature vectors. In this embodiment, the fine-grained feature vector, the medium-grained feature vector, and the coarse-grained feature vector are respectively denoted as... , as well as ; It should be noted that each feature vector corresponds to the mean, variance, standard deviation, peak value, trough value, range, linear fitting slope, trend correlation coefficient, coefficient of variation, fluctuation frequency, and fluctuation amplitude at a corresponding granularity, and is arranged in a fixed order to ensure that the vector dimensions are consistent.
[0040] For example, if the time window lengths of the time granular level are 5s, 15s, and 30s respectively, then the data of the multi-granularity monitoring feature dataset corresponding to the time granular level with a time window length of 5s are denoted as fine-grained feature vector, medium-grained feature vector, and coarse-grained feature vector, and so on.
[0041] In practical applications, the specific steps for performing cross-granularity correlation analysis on the multi-granularity monitoring feature dataset are as follows: B1: Calculate the first cross-granularity correlation coefficient based on the fine-grained feature vector and the medium-grained feature vector; B2: Calculate the second cross-granularity correlation coefficient based on the fine-grained feature vector and the coarse-grained feature vector; B3: Based on the first cross-granularity correlation coefficient and the second cross-granularity correlation coefficient, obtain the global fusion feature difference for the corresponding acquisition period.
[0042] In steps B1 and B2, the cosine similarity is calculated using the cosine similarity method to determine the cosine similarity between corresponding feature vectors in the fine-grained and medium-grained feature vectors, and between the fine-grained and coarse-grained feature vectors. The cosine similarity obtained from the fine-grained and medium-grained feature vectors is recorded as the first cross-granularity correlation coefficient. The cosine similarity obtained from the fine-grained and coarse-grained feature vectors is denoted as the second cross-granularity correlation coefficient. .
[0043] It should be noted that in step B3, the global fusion feature difference corresponding to the acquisition period... .
[0044] Preset global fusion feature difference threshold; If the global fusion feature difference If the difference is less than or equal to the global fusion feature threshold, it indicates that the chiller unit is operating normally during the current acquisition period. If the global fusion feature difference If the difference is greater than the global fusion feature difference threshold, it indicates that the operating status of the refrigeration unit in the current acquisition period is abnormal. The multi-source operating status data and cold storage environment data of the corresponding acquisition period are then obtained. Based on the multi-source operating status data and cold storage environment data of the corresponding acquisition period, the operating status of the corresponding equipment is specifically analyzed. If there is an abnormality, it will be marked and a corresponding abnormal equipment status set will be formed; otherwise, the corresponding equipment operating status is normal.
[0045] Step S4: Based on LSTM and according to the global fusion feature difference, obtain the ultra-short-term state prediction model of the cold storage refrigeration unit, and according to the ultra-short-term state prediction model, obtain the operating status dataset of each refrigeration unit within the ultra-short monitoring period.
[0046] It should be noted that LSTM (Long Short-Term Memory) is an improved recurrent neural network that effectively solves the gradient vanishing and gradient exploding problems of traditional recurrent neural networks through a gating mechanism, making it suitable for predictive analysis of time series data.
[0047] In practical applications, the global fusion feature difference and the global fusion feature difference threshold are obtained from several historical acquisition cycles. The global fusion feature differences and global fusion feature difference thresholds from several historical acquisition periods are grouped and labeled, denoted as follows: It is a natural number; Pick The global fusion feature difference and the global fusion feature difference threshold of the historical acquisition period are used as the sample set, and Less than The natural number is used as the test set, and the global fusion feature difference and global fusion feature difference threshold of the remaining historical collection periods are used as the test set. The training sample set is formed based on the sample set and the test set.
[0048] A standard state prediction model is constructed based on LSTM. The training sample set is input into the standard state prediction model for training. When the loss function of the sample set is consistent with the trend of the loss function of the test set and the loss does not decrease for 10 consecutive rounds, the training is stopped and the current optimal model parameters are saved. The trained standard state prediction model is then recorded as the ultra-short-term state prediction model, and the operating status of the refrigeration unit is output.
[0049] It should be noted that the LSTM used in this application is a general-purpose technology with the same architecture as conventional LSTM. Specific parameter settings can be further configured according to actual needs, meeting both accuracy and precision requirements. A three-layer architecture is adopted, including an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is consistent with the dimension of all feature particles in the feature particle hierarchy, and the ReLU activation function is used. The hidden layer contains two LSTM layers: the first layer has 128 neurons, and the second layer has 64 neurons, both employing Dropout with a Dropout rate of 0.2 to prevent overfitting. The output layer is a fully connected layer with one neuron, using a linear activation function, and outputs the operating status of the refrigeration unit. To further clarify, the model uses mean squared error (MSE) as the loss function, Adam optimizer as the optimizer, learning rate of 0.001, and 100 iterations.
[0050] In practical applications, the global fusion feature difference of the current acquisition period is input into the ultra-short-term state prediction model, which outputs the operating status of the refrigeration unit in the current acquisition period. If the operating status of the refrigeration unit in the current acquisition period is abnormal, multi-source operating status data and cold storage environment data for the corresponding acquisition period are acquired. Based on the multi-source operating status data and cold storage environment data for the corresponding acquisition period, the operating status of the corresponding equipment in the refrigeration unit is analyzed. If the corresponding equipment is abnormal, it will be marked as an important item, and a set of abnormal equipment statuses for the refrigeration unit in the current acquisition period will be formed. If the corresponding equipment is operating normally, it will be marked as normal, and a set of normal equipment statuses for the refrigeration unit in the current acquisition period will be formed. The abnormal equipment status set and the normal equipment status set are combined to form an operating status dataset; based on the operating status dataset, a standardized remote monitoring report is generated; the remote monitoring report includes, but is not limited to, basic information of the chiller unit, power curves within the monitoring period, distribution of operating status levels, and targeted adjustment suggestions.
[0051] Example 2: like Figure 2 As shown, a remote monitoring platform system for cold storage refrigeration units includes, but is not limited to, a data acquisition module, a data preprocessing module, a particle feature processing module, a cross-particle size analysis module, and a state prediction module. The data acquisition module is used to collect multi-source operating status data of the cold storage refrigeration unit and cold storage environmental data; The data preprocessing module is used to preprocess multi-source operating status data and cold storage environment data to obtain corresponding standardized operating status data and standardized cold storage environment data. The granular feature processing module, based on granular computing theory, performs granular feature processing on standardized operating status data and standardized cold storage environment data to obtain corresponding multi-granularity monitoring feature datasets. The cross-granularity analysis module is used to perform cross-granularity correlation analysis on multi-granularity monitoring feature datasets to obtain the corresponding global fusion feature differences; The state prediction module, based on LSTM, obtains an ultra-short-term state prediction model for the cold storage refrigeration units according to the global fusion feature difference, and obtains the operating state dataset of each refrigeration unit within an ultra-short monitoring period based on the ultra-short-term state prediction model.
[0052] Optionally, in this embodiment, those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.
[0053] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0054] If the integrated units in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in the aforementioned computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause one or more electronic devices to execute all or part of the steps of the methods described in the various embodiments of this application.
[0055] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0056] In the several embodiments provided in this application, it should be understood that the disclosed application can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.
[0057] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0058] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0059] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.
Claims
1. A method for remote platform monitoring of cold storage refrigeration units, characterized in that, The method includes: Collect multi-source operating status data of the cold storage refrigeration unit and cold storage environment data, and perform data preprocessing on the multi-source operating status data and cold storage environment data to obtain corresponding standardized operating status data and standardized cold storage environment data. Based on granular computing theory, granular feature processing is performed on standardized operating status data and standardized cold storage environment data to obtain corresponding multi-granularity monitoring feature datasets. Perform cross-granularity correlation analysis on the multi-granularity monitoring feature dataset to obtain the corresponding global fusion feature difference; Based on LSTM and according to the global fusion feature difference, an ultra-short-term state prediction model for cold storage refrigeration units is obtained. Based on the ultra-short-term state prediction model, the operating status dataset of each refrigeration unit within an ultra-short monitoring period is obtained.
2. The remote platform monitoring method for a cold storage refrigeration unit according to claim 1, characterized in that, Based on granular computing theory, the process of granular feature processing for standardized operational status data and standardized cold storage environment data includes: The standardized operating status data and standardized cold storage environment data are divided into feature granules to obtain the corresponding feature granule levels; Each feature particle is divided into time dimensions using a fixed time window. The length of the time window is configurable. Based on the time window length, the time particle hierarchy corresponding to the time window length is obtained. Based on the time granularity level and the feature granularity level, granular feature processing is performed on standardized operating status data and standardized cold storage environment data to obtain the corresponding multi-granularity monitoring feature dataset.
3. The remote platform monitoring method for a cold storage refrigeration unit according to claim 2, characterized in that, The process of performing granular feature processing on standardized operational status data and standardized cold storage environment data based on time granularity and feature granularity includes: Determine the time granularity level and the feature granularity level, using the feature granularity level-time granularity level as a unique identifier; group the standardized operating status data and standardized cold storage environment data according to the unique identifier. Based on the unique identifier, statistical feature processing, trend feature processing, and fluctuation feature processing are performed on each group of data to obtain the statistical feature set, trend feature set, and fluctuation feature set. The statistical feature set includes the mean, variance, standard deviation, peak value, trough value, and range of the corresponding data; the trend feature set includes the linear fit slope and trend correlation coefficient of the corresponding data; and the fluctuation feature set includes the coefficient of variation, fluctuation frequency, and fluctuation amplitude of the corresponding data.
4. The remote platform monitoring method for a cold storage refrigeration unit according to claim 3, characterized in that, The process of performing trend feature processing on each group of data includes: During the acquisition period, the starting time of the time window of the time particle in the corresponding time particle level is set to 0, and the time window length is incremented according to the corresponding time particle level to form time series data. Linear regression analysis of time series data is performed using the least squares method to obtain the linear fit slope; the corresponding trend correlation coefficient is obtained based on the Pearson correlation coefficient.
5. The remote platform monitoring method for a cold storage refrigeration unit according to claim 4, characterized in that, The process of performing fluctuation feature processing on each group of data includes: Calculate the coefficient of variation of the corresponding data based on the standard deviation and mean obtained from statistical feature processing; Traverse the time-series data within the information granules, compare the magnitude of two adjacent data points with the mean of the corresponding data points, and count the number of effective crossings; obtain the fluctuation frequency based on the number of effective crossings and the corresponding time window length; obtain the fluctuation amplitude based on the mean of the corresponding data and the data at the actual collection time point.
6. The remote platform monitoring method for a cold storage refrigeration unit according to claim 5, characterized in that, The process of performing cross-granularity correlation analysis on the multi-granularity monitoring feature dataset includes: The multi-granularity monitoring feature dataset is divided into fine-grained feature vectors, medium-grained feature vectors, and coarse-grained feature vectors. Calculate the first cross-granularity correlation coefficient based on the fine-grained feature vector and the medium-grained feature vector; calculate the second cross-granularity correlation coefficient based on the fine-grained feature vector and the coarse-grained feature vector. Based on the first cross-granularity correlation coefficient and the second cross-granularity correlation coefficient, the global fusion feature difference for the corresponding acquisition period is obtained; Preset global fusion feature difference threshold; If the global fusion feature difference is less than or equal to the global fusion feature difference threshold, it indicates that the chiller unit is operating normally during the current acquisition period. If the difference in global fusion features is greater than the threshold for global fusion feature difference, it indicates that the operating status of the refrigeration unit in the current acquisition period is abnormal. The multi-source operating status data and cold storage environment data of the corresponding acquisition period are then obtained. Based on the multi-source operating status data and cold storage environment data of the corresponding acquisition period, the operating status of the corresponding equipment is specifically analyzed. If there is an abnormality, it will be marked and a corresponding set of abnormal equipment statuses will be formed; otherwise, the corresponding equipment operating status is normal.
7. The remote platform monitoring method for a cold storage refrigeration unit according to claim 6, characterized in that, The process of constructing an ultra-short-term state prediction model includes: Obtain the global fusion feature difference and the global fusion feature difference threshold for several historical acquisition periods; The global fusion feature differences and global fusion feature difference thresholds from several historical acquisition periods are grouped and labeled, denoted as follows: It is a natural number; Pick The global fusion feature difference and the global fusion feature difference threshold of the historical acquisition period are used as the sample set, and Less than The natural number, the global fusion feature difference and the global fusion feature difference threshold of the remaining historical collection periods are used as the test set, and the training sample set is formed according to the sample set and the test set; A standard state prediction model is constructed based on LSTM. The training sample set is input into the standard state prediction model for training, and the trained standard state prediction model is denoted as the ultra-short-term state prediction model.
8. A remote platform monitoring system for cold storage refrigeration units, implementing the remote platform monitoring method for cold storage refrigeration units as described in any one of claims 1 to 7, characterized in that, include: The module includes a data acquisition module, a data preprocessing module, a particle feature processing module, a cross-particle size analysis module, and a state prediction module. The data acquisition module is used to collect multi-source operating status data of the cold storage refrigeration unit and cold storage environmental data; The data preprocessing module is used to preprocess multi-source operating status data and cold storage environment data to obtain corresponding standardized operating status data and standardized cold storage environment data. The granular feature processing module, based on granular computing theory, performs granular feature processing on standardized operating status data and standardized cold storage environment data to obtain corresponding multi-granularity monitoring feature datasets. The cross-granularity analysis module is used to perform cross-granularity correlation analysis on multi-granularity monitoring feature datasets to obtain the corresponding global fusion feature differences; The state prediction module, based on LSTM, obtains an ultra-short-term state prediction model for the cold storage refrigeration units according to the global fusion feature difference, and obtains the operating state dataset of each refrigeration unit within an ultra-short monitoring period based on the ultra-short-term state prediction model.