Commercial display cabinet remote intelligent operation and maintenance management system based on big data

Abstract

The invention relates to the technical field of remote monitoring, in particular to a commercial display cabinet remote intelligent operation and maintenance management system based on big data, which comprises an edge waveform acquisition module, a starting energy resolving module, a same-pressure working condition clustering module, a group baseline generation module and a wear outlier judgment module. According to the invention, through constructing an edge feature extraction and cloud voltage coupling clustering cooperative processing flow, a dynamic voltage impedance relative evaluation system is constructed by using wide-area big data, starting energy required for overcoming static friction force and rotor inertia is quantized, and group baselines of similar display cabinets under similar voltage working conditions are combined, so that the relative evaluation of the dynamic voltage impedance is realized. The starting energy of the target display cabinet is transversely compared with the group level, accurate relative quantification and early warning are carried out on abrasion of mechanical parts in the compressor under the condition that the influence of power grid quality difference is eliminated, the fault diagnosis accuracy of the display cabinet in a voltage unstable area is improved, false overload alarm caused by low voltage is avoided, and the working efficiency is improved. And passive maintenance is converted into predictive maintenance.

Description

Technical Field

[0001] This invention relates to the field of remote monitoring technology, and in particular to a remote intelligent operation and maintenance management system for commercial display cabinets based on big data. Background Technology

[0002] The field of remote monitoring technology involves a technical system that utilizes wired or wireless communication networks to collect, transmit, and centrally monitor the operating status and environmental parameters of remote display cabinets in real time. The core aspects of this field encompass data acquisition from the front-end sensing display cabinet, signal transmission at the network layer, and data presentation and command control at the monitoring center, forming a display cabinet management architecture that transcends geographical limitations. Traditional commercial display cabinet remote intelligent operation and maintenance management systems refer to systems that maintain and manage the electrical parameters and cooling / heating performance of merchandise display cabinets in scenarios such as supermarkets and convenience stores. Typically, a fixed temperature range and compressor current threshold are preset in the embedded controller of the display cabinet. Real-time operating data is collected using temperature sensors and current transformers. When the collected values ​​exceed the preset threshold range, the controller triggers a local audible and visual alarm and controls the communication module to send a specific fault code to the central server. The server records the alarm log, and a dispatcher calls maintenance personnel with a multimeter and pressure gauge to check the display cabinet's condition on-site and perform manual troubleshooting and component replacement.

[0003] Existing technologies rely solely on preset fixed current thresholds to diagnose compressor faults, ignoring the interference of grid voltage fluctuations in different regions on motor starting characteristics. This leads to false alarms in areas with unstable voltage, and makes it impossible to accurately identify abnormal mechanical impedance during the early wear stage before the compressor completely seizes up. It can only respond passively based on fault codes after the fact, lacking quantitative assessment methods for the sub-health status of display cabinets. This makes it difficult to reduce the risk of merchandise loss caused by display cabinet downtime. Furthermore, relying on manual dispatch to notify maintenance personnel to go to the site for troubleshooting is inefficient, increases operation and maintenance costs, and cannot achieve refined control of the display cabinet's operating status under different voltage environments. Summary of the Invention

[0004] To address the shortcomings of existing technologies, such as ignoring the interference of grid voltage fluctuations in different regions on motor starting characteristics, leading to false alarms in unstable voltage areas, failing to accurately identify abnormal mechanical impedance during the early wear stage before the compressor completely seizes up, relying solely on reactive fault codes, lacking quantitative assessment methods for the sub-health state of display cabinets, making it difficult to reduce the risk of merchandise loss due to display cabinet downtime, and relying on manual dispatching of maintenance personnel for on-site troubleshooting is inefficient and increases operation and maintenance costs, and failing to achieve refined management of display cabinet operating status under different voltage environments, this invention provides a remote intelligent operation and maintenance management system for commercial display cabinets based on big data. The technical solution is as follows: On the one hand, it provides a remote intelligent operation and maintenance management system for commercial display cabinets based on big data. This system includes: The edge waveform acquisition module monitors the closing action status of the relay control signal of the terminal compressor of the commercial display cabinet, acquires transient current and real-time voltage waveform data, and generates a startup transient waveform data package; The startup energy calculation module performs integration on the transient current waveform data packet to obtain the startup energy, performs root mean square calculation on the real-time voltage waveform data to obtain the voltage RMS value, and associates the startup energy and voltage RMS value to generate startup feature data pairs. The same-pressure working condition clustering module parses the model identifier, retrieves active data with the model identifier, sets the voltage deviation range based on the effective voltage value, filters active data that meet the voltage deviation range to extract the start-up energy, and establishes a data cluster of display cabinets of the same type and under the same pressure. The group baseline generation module sorts the starting energy in the data clusters of the same type of display cabinets under the same pressure, calculates the median index and interquartile range index, and constructs a group impedance distribution benchmark. The wear outlier determination module calculates the difference between the starting energy and the median index in the group impedance distribution benchmark, divides the median index difference by the interquartile range index to obtain the deviation index, compares the deviation index with the preset abnormal threshold, and generates a wear warning determination result for the display cabinet.

[0005] As a further aspect of the present invention, the startup transient waveform data packet includes a collection trigger timestamp, a dual-channel sampling point array, and sensor calibration parameters; the startup characteristic data pair includes mechanical work characterization values, power grid supply capacity characterization values, and time-series correlation tags; the same-voltage and same-type display cabinet data cluster includes a display cabinet identification set of the same model, an approximate voltage operating condition record set, and a startup energy value sequence; the group impedance distribution benchmark includes a group energy median line, impedance dispersion range, and statistical sample capacity identifier; and the display cabinet wear warning judgment result includes a mechanical impedance anomaly level, a display cabinet maintenance suggestion code, and a fault occurrence probability assessment value.

[0006] As a further aspect of the present invention, the edge waveform acquisition module includes: The trigger signal capture submodule monitors the closing action status of the compressor relay control signal of the commercial display cabinet terminal. When the rising state of the signal representing the closing action is detected, the clock is locked, real-time time point data is obtained as a reference, and the time point data is logically associated with the hardware start request signal to generate a data acquisition synchronization trigger command. The dual-channel synchronous acquisition submodule responds to the acquisition synchronization trigger command, synchronously activates the current transformer and voltage sensor connected to the compressor power supply line, continuously reads the analog signal within a fixed time window during the startup process according to the preset sampling frequency, converts the analog signal into digital quantized data through the analog-to-digital converter and establishes the corresponding time index, and generates an electrical transient waveform sequence. The data packet encapsulation submodule calls the electrical transient waveform sequence, performs time-domain synchronization alignment operation on the current change trajectory and voltage fluctuation trajectory according to the time index, extracts the sampling frequency and display cabinet identifier of the waveform sequence as header metadata, and integrates the aligned dual-channel waveform data with the header metadata through binary serialization to construct the startup transient waveform data packet.

[0007] As a further aspect of the present invention, the startup energy calculation module includes: The current integration operation submodule calls the startup transient waveform data packet, deconstructs the transient current waveform data sequence and timestamp index, performs a product operation on the instantaneous current amplitude of each discrete sampling point and the time interval between adjacent sampling points, performs an accumulation and summation operation on the product terms within the preset startup time window, quantifies the total integral of current with time during the process of the compressor rotor accelerating from a stationary state to the rated speed, and generates the startup charge accumulation value. The voltage root mean square calculation submodule responds to the calculation completion signal of the accumulated starting charge value, extracts real-time grid voltage waveform data synchronized with the energy value time span, performs a square operation on each discrete voltage sampling point in the real-time grid voltage waveform data sequence, calculates the arithmetic mean of the square values ​​and performs square root processing on the mean, obtains the voltage physical quantity characterizing the real-time load capacity of the power supply network at the moment of startup, and generates the effective value of grid voltage. The feature association packaging submodule establishes a logical mapping relationship between the two values ​​based on the effective value of the grid voltage and the unique identifier of the display cabinet startup event. It marks the startup charge accumulation value as a mechanical load characteristic index and the effective value of the grid voltage as an input voltage condition index. According to the preset communication protocol format, the indexes are combined into a structured data unit with spatiotemporal correlation to generate startup feature data pairs.

[0008] As a further aspect of the present invention, the same-pressure condition clustering module includes: The same type data retrieval submodule calls the startup feature data pair, parses the target display cabinet model identifier information, performs condition matching retrieval for display cabinet records across the entire network, locks records with the same model identifier and valid upload behavior within a preset recent time window, extracts the corresponding running status parameters and identity index information of the records, and generates a dataset of active display cabinets of the same type. The voltage window setting submodule calls the grid voltage effective value in the startup feature data pair as the reference point, performs addition and subtraction operations according to the preset voltage fluctuation tolerance parameter, calculates the reference voltage rise and fall limit values ​​respectively, and generates the same voltage screening deviation range. The working condition alignment clustering submodule calls the same pressure screening deviation range, extracts the voltage value corresponding to the record and performs numerical inclusion relationship determination with the same pressure screening deviation range, retains specific records whose voltage values ​​fall within the range, separates the corresponding start-up energy values ​​from the specific records and aggregates them into a single-dimensional numerical set to establish a same pressure and same type display cabinet data cluster.

[0009] As a further aspect of the present invention, an inverted index structure is established for the display cabinet records across the entire network, and the candidate data whose timestamp attributes fall within a continuous time interval is located through the inverted index structure; The system checks whether the candidate data includes a complete starting current waveform file and the corresponding compressor operating impedance data, and only records containing complete data are marked as having valid upload behavior.

[0010] As a further aspect of the present invention, the population baseline generation module includes: The numerical sorting and sorting submodule calls the data cluster of the same pressure and type of display cabinet, traverses the starting energy values ​​of the independent display cabinets, performs a numerical comparison operation, rearranges the energy data according to monotonically increasing logic, establishes the correspondence between numerical value and sequence position, removes the time dimension index interference in the original data, and generates an ascending energy value list. The distribution feature calculation submodule calls the ascending energy value list, locates the value point in the middle position according to the total length of the ascending energy value list as the central trend characterization, locates the value points in the quarter position and three-quarter position of the ascending energy value list respectively, performs subtraction operation on the two quantile values ​​to obtain the difference reflecting the data dispersion width, combines the central position value and the dispersion width difference to form mathematical parameters for quantifying the population distribution state, and generates a distribution statistical feature set; The benchmark model construction submodule, based on the aforementioned distribution statistical feature set, maps the median value to the standard mechanical impedance reference line of the model display cabinet under a specific voltage, maps the interquartile range to the allowable reasonable fluctuation range width, and constructs a multi-dimensional benchmark data structure by combining real-time sample capacity information to generate a group impedance distribution benchmark.

[0011] As a further aspect of the present invention, the process of constructing a multi-dimensional benchmark data structure by combining real-time sample capacity information is as follows: calling the logarithmic calculation logic to calculate the logarithm of the list length value with base ten, and multiplying the logarithm by the preset confidence weight coefficient to obtain the sample confidence score. A structured storage space is allocated in memory, and a standard mechanical impedance reference line, the allowable reasonable fluctuation range width, and the sample confidence score are written in sequence. The minimum and maximum values ​​in the ascending energy value list are written into the structured storage space as boundary constraint parameters to form the group impedance distribution benchmark including five-dimensional data fields.

[0012] As a further aspect of the present invention, the wear outlier determination module includes: The impedance difference quantification submodule uses the median value index stored in the group impedance distribution benchmark as a reference anchor point, and combines the startup charge accumulation value recorded in the startup characteristic data to perform a subtraction algebra operation on the median value index and the startup charge accumulation value to calculate the numerical difference between the real-time energy consumption of the target display cabinet in the startup event and the central trend of the same type of display cabinet group, and generates an energy benchmark deviation value. The outlier normalization submodule, based on the energy reference deviation value, calls the interquartile range index recorded in the population impedance distribution reference as a scale to measure the degree of data dispersion. It performs a division operation with the energy reference deviation value as the numerator and the interquartile range index as the denominator, transforming the absolute difference in the physical energy dimension into a relative distance in the statistical dimension, and generating an impedance deviation coefficient. The wear status determination submodule uses the impedance deviation coefficient to read the preset mechanical wear anomaly determination threshold, performs a numerical comparison operation between the impedance deviation coefficient and the anomaly determination threshold, and triggers the anomaly marking logic if the impedance deviation coefficient is greater than the anomaly determination threshold. Based on the magnitude of the impedance deviation coefficient exceeding the anomaly determination threshold, the corresponding risk severity level is matched, and a wear warning determination result for the display cabinet is generated.

[0013] As a further aspect of the present invention, the process of performing subtraction algebraic operation on the median index and the cumulative value of the starting charge is specifically as follows: read the cumulative value of the starting charge recorded in the starting characteristic data as the minuend, read the median index stored in the group impedance distribution reference as the subtrahend, and perform double-precision floating-point subtraction to obtain the original difference value. A unidirectional wear drift filtering logic is constructed to determine whether the original difference value is greater than 0. When the original difference value is greater than 0, it indicates that there is an increase in positive impedance in the display cabinet, and the original difference value is assigned as the energy reference deviation value. When the original difference value is less than or equal to 0, it indicates that the display case is in a low impedance operating state. The energy reference deviation value is forcibly set to 0 to eliminate negative deviations caused by non-wear factors.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By focusing on eliminating environmental interference, a collaborative processing flow of edge feature extraction and cloud-based voltage coupling clustering is constructed. A dynamic voltage impedance relative evaluation system is built using wide-area big data. The waveform data at the moment of compressor startup is monitored and integral calculations are performed to quantify the startup energy required to overcome static friction and rotor inertia. Combined with the group baseline of similar display cabinets under similar voltage conditions, the startup energy of the target display cabinet is compared with the group level. Under the condition of eliminating the influence of power grid quality differences, the wear of internal mechanical components of the compressor is accurately relative quantified and early warning is provided. This improves the fault diagnosis accuracy of display cabinets in voltage unstable areas, avoids false overload alarms caused by low voltage, and transforms passive maintenance into predictive maintenance. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A schematic diagram of the system provided by the present invention; Figure 2 This is a schematic diagram of the system framework of the present invention; Figure 3 This is a flowchart of the edge waveform acquisition module in this invention; Figure 4 This is a flowchart of the energy calculation module startup process in this invention; Figure 5 This is a flowchart of the same-pressure working condition clustering module in this invention; Figure 6 This is a flowchart of the population baseline generation module in this invention; Figure 7 This is a flowchart of the wear outlier determination module in this invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] This invention provides a remote intelligent operation and maintenance management system for commercial display cabinets based on big data, such as... Figure 1-2 The diagram shown illustrates a remote intelligent operation and maintenance management system for commercial display cabinets based on big data. This system includes: The edge waveform acquisition module monitors the closing status of the control signal of the compressor relay in the commercial display cabinet terminal, triggers the current transformer and voltage sensor to perform synchronous acquisition operations, obtains transient current waveform data and real-time grid voltage waveform data within a preset time window during the compressor startup process, and generates a startup transient waveform data package; The startup energy calculation module calls the startup transient waveform data package, performs integral calculation on the transient current waveform data, performs root mean square calculation on the real-time grid voltage waveform data, and associates and packages the calculated startup energy value with the effective voltage value to generate startup feature data pairs. The same-pressure working condition clustering module analyzes the target display cabinet model identification information, retrieves active display cabinet data with the same model identification and uploaded records within a preset period, sets the voltage deviation range with the effective voltage value of the start-up feature data as the center, filters display cabinet records whose effective voltage values ​​fall within the voltage deviation range and extracts the start-up energy value, and establishes a data cluster of display cabinets of the same type under the same pressure. The group baseline generation module performs a data sorting operation based on the data cluster of display cabinets of the same pressure and type, arranges the display cabinet start-up energy values ​​in ascending order, calculates the median index and interquartile range index at the center of the data sequence, and constructs a group impedance distribution benchmark. The wear outlier determination module calculates the difference between the starting energy value and the median index in the group impedance distribution benchmark. The difference is divided by the interquartile range index to obtain the deviation index. The deviation index is compared with the preset anomaly determination standard to generate the wear warning determination result of the display cabinet. The startup transient waveform data package includes the acquisition trigger timestamp, dual-channel sampling point array, and sensor calibration parameters. The startup characteristic data pair includes mechanical work characterization value, power grid supply capacity characterization value, and time-series correlation label. The same voltage and type of display cabinet data cluster includes the same model display cabinet identification set, approximate voltage operating condition record set, and startup energy value sequence. The group impedance distribution benchmark includes the group energy median line, impedance dispersion range, and statistical sample capacity identifier. The display cabinet wear warning judgment result includes the mechanical impedance abnormality level, display cabinet maintenance suggestion code, and fault occurrence probability assessment value.

[0019] Specifically, such as Figure 2 , 3 As shown, the edge waveform acquisition module includes: The trigger signal capture submodule monitors the closing action status of the compressor relay control signal of the commercial display cabinet terminal. When the rising state of the signal representing the closing action is detected, the clock is locked, real-time time point data is obtained as a reference, and the time point data is logically associated with the hardware start request signal to generate a data acquisition synchronization trigger command. As an edge sensing front-end in a cloud-edge collaborative architecture, this submodule is physically connected to the control coil input of the compressor relay via a high-frequency optocoupler isolation circuit to detect real-time level changes on the control line. Internally, this submodule is equipped with a high-precision crystal oscillator clock source with a frequency of 10 MHz. When the control line voltage jumps from 0 volts to 24 volts or a preset high-level logic threshold, the output of the optocoupler isolation circuit generates a corresponding digital pulse signal. The interrupt controller inside the submodule captures the rising edge of this signal and immediately triggers the hardware latch, locking the current count value of the crystal oscillator clock source into a register. This acquires real-time time data accurate to the microsecond level as a reference timestamp. Subsequently, the submodule reads the hardware start request signal status bit issued by the main control chip from the system memory to verify whether the relay closing action belongs to the expected normal start-up process. If the status bit verification passes, the submodule writes the locked timestamp data into the header field of the instruction buffer and writes the preset sensor activation mask and sampling channel configuration parameters into the instruction payload field. It then sends a data acquisition synchronization trigger instruction consisting of a 32-bit timestamp and a 16-bit control word to the subsequent processing unit via the internal high-speed bus.

[0020] The dual-channel synchronous acquisition submodule responds to the acquisition synchronization trigger command, synchronously activates the current transformer and voltage sensor connected to the compressor power supply line, continuously reads the analog signal within a fixed time window during the startup process according to the preset sampling frequency, converts the analog signal into digital quantized data through the analog-to-digital converter and establishes the corresponding time index, and generates an electrical transient waveform sequence. Upon receiving the acquisition synchronization trigger command on the bus, the module executes a local high-frequency edge acquisition task. Immediately after receiving the command, it parses the channel configuration parameters and simultaneously sends enable signals to the Hall effect current transformer connected to the compressor's L-phase power supply line and the resistance divider voltage sensor connected in parallel to the power input within 10 microseconds, synchronously activating these two analog signal acquisition front-ends. The 16-bit successive approximation analog-to-digital converter integrated within the submodule triggers a conversion action every 50 microseconds based on a preset 20 kHz sampling frequency, within a fixed 500 millisecond time window set during the compressor startup process. The analog voltage signals output from the current transformer and the voltage sensor are continuously read and converted, resulting in a total of 10,000 discrete current sampling points and 10,000 discrete voltage sampling points. The analog-to-digital converter maps the amplitude of the analog signal obtained from each conversion to digital quantized data between 0 and 65535, and uses a direct memory access controller to write the data sequentially to a specific address segment of the dual-port random access memory. At the same time, each data point is associated with a time offset relative to the reference timestamp, establishing an electrical transient waveform sequence with strict timing correspondence.

[0021] The data packet encapsulation submodule calls the electrical transient waveform sequence, performs time-domain synchronization alignment operation on the current change trajectory and voltage fluctuation trajectory according to the time index, extracts the sampling frequency and display cabinet identifier of the waveform sequence as header metadata, and integrates the aligned dual-channel waveform data with the header metadata through binary serialization to construct the startup transient waveform data packet; To prepare for efficient data transfer to the cloud, the submodule directly accesses the dual-port random access memory via memory mapping to read the stored electrical transient waveform sequence. First, the submodule checks the length and time index of the current and voltage data arrays. For minor time deviations caused by clock jitter or transmission delay, phase interpolation compensation is performed on the current data, using the zero-crossing point of the voltage data as a reference. Time-domain synchronization alignment is then performed to ensure that the current and voltage values ​​under the same index correspond to the same physical moment. Subsequently, the submodule reads the current sampling frequency value of 20000 and the 16-byte unique identifier of the display cabinet stored in the read-only memory from the display cabinet configuration register, filling this information into the header metadata area defined in the data packet. Using a low-byte order format, the submodule performs binary serialization and concatenates the aligned 10000 current and 10000 voltage data points to form a continuous load data block. A CRC32 cyclic redundancy check code is calculated and appended to the end of the data block, constructing a startup transient waveform data packet with a total length of approximately 40 kilobytes, containing complete header information and waveform load.

[0022] Specifically, such as Figure 2 , 4 As shown, starting the energy calculation module includes: The current integration operation submodule calls the startup transient waveform data packet, deconstructs the transient current waveform data sequence and timestamp index, performs a product operation on the instantaneous current amplitude of each discrete sampling point and the time interval between adjacent sampling points, performs an accumulation and summation operation on the product terms within the preset startup time window, quantifies the total amount of current integral over time during the process of the compressor rotor accelerating from a stationary state to the rated speed, and generates the startup charge accumulation value. Utilizing edge computing resources for feature extraction, the module receives and decompresses the transient waveform data packet, restoring the binary payload to a transient current waveform data sequence containing 10,000 sampling points and corresponding timestamp indices. The submodule initializes the integrator accumulator to 0 and begins traversing the current data sequence. For each discrete sampling point in the sequence, the submodule reads the instantaneous current amplitude at that point and calculates the time difference between the timestamp of that sampling point and the timestamp of the next adjacent sampling point. For example, at a 20 kHz sampling rate, this time interval is fixed at 0.00005 seconds. The submodule performs a product operation, comparing the instantaneous current amplitude with the 0.00005-second time interval. Multiply by the area of ​​the current rectangle within this tiny time segment, and then accumulate the product into the integrator. This accumulation process continues to cover the entire preset 500-millisecond startup time window. Through this discretized Riemann sum calculation method, the continuous integration process is simulated, quantifying the total amount of current integrated over time during the process of the compressor rotor overcoming the maximum static friction and accelerating to the rated speed from a stationary state. That is, the cumulative effect of the startup impact current in the time domain, and finally the value in the accumulator is output. If the average impact current during a startup process is 15 amperes and the duration is 0.2 seconds, after point-by-point integration calculation, the generated startup charge accumulation value is approximately 3 coulombs.

[0023] The voltage root mean square calculation submodule responds to the signal indicating that the calculation of the accumulated charge value has been completed, extracts real-time grid voltage waveform data synchronized with the energy value time span, performs a square operation on each discrete voltage sampling point in the real-time grid voltage waveform data sequence, calculates the arithmetic mean of the squared values ​​and performs square root processing on the mean, obtains the voltage physical quantity characterizing the real-time load capacity of the power supply network at the moment of startup, and generates the effective value of the grid voltage. After completing data cleaning and preliminary calculations at the edge, and detecting the status flag indicating the completion of the startup energy calculation, the submodule immediately locks onto the time period in the startup transient waveform data packet that completely overlaps with the startup energy integration interval. It extracts the real-time grid voltage waveform data within this time period, which also contains 10,000 voltage sampling points corresponding to a 500-millisecond window. The submodule initializes a sum-of-squares register, iterates through these 10,000 discrete voltage sampling points, performs a self-multiplication operation (square operation) on the voltage amplitude of each sampling point, and accumulates all calculated square values ​​into the sum-of-squares register. After all points have been traversed, the submodule divides the total value in the sum-of-squares register by the total number of sampling points, 10,000, to calculate the arithmetic mean of the voltage squares. Finally, the submodule calls the floating-point arithmetic unit to perform an arithmetic square root operation on this arithmetic mean, obtaining the voltage physical quantity that can equivalently characterize the AC power supply network's ability to perform work at the moment of startup, generating the effective value of the grid voltage. For example, if the peak value of the acquired voltage waveform is 311 volts, after the above square, average, and square root calculation process, the calculated effective value of the grid voltage is 220 volts.

[0024] The feature association packaging submodule establishes a logical mapping relationship between two values ​​based on the effective value of the grid voltage and the unique identifier of the display cabinet start event. It marks the start charge accumulation value as a mechanical load characteristic index and the effective value of the grid voltage as an input voltage condition index. According to the preset communication protocol format, the indexes are combined into a structured data unit with spatiotemporal correlation to generate start feature data pairs. A standardized payload conforming to the cloud-edge collaborative communication protocol is constructed. A new structure space is allocated in memory, and the unique event identifier generated when the display cabinet startup event is triggered is read and written into the index field of the structure as the primary key for subsequent data traceability. Subsequently, the submodule maps the startup charge accumulation value calculated in the previous steps into the mechanical load characteristic field of the structure, and maps the grid voltage effective value into the input voltage condition field of the structure, establishing a one-to-one logical mapping relationship between the two values. The submodule converts the above structure into a string form key-value pair combination according to the JSON data format defined by the MQTT communication protocol. For example, the startup energy is marked as "start_energy" and the voltage effective value is marked as "grid_voltage". The display cabinet ID and upload timestamp are added and combined into a structured data unit with spatiotemporal correlation to generate startup feature data pairs.

[0025] Specifically, such as Figure 2 , 5 As shown, the same-pressure condition clustering module includes: The same-type data retrieval submodule calls the start feature data pair, parses the target display cabinet model identifier information, performs condition matching retrieval on display cabinet records across the entire network, locks records with the same model identifier and valid upload behavior within a preset recent time window, extracts the corresponding running status parameters and identity index information of the records, and generates a dataset of active display cabinets of the same type. Leveraging a massive cloud-based database for wide-area retrieval, the received startup feature data is parsed using JSON to extract the target display cabinet model identifier information, such as "Model-850L". Subsequently, the submodule connects to the cloud-based database via an encrypted channel to construct a structured query statement with two core filtering conditions: firstly, the display cabinet model field must strictly match "Model-850L"; secondly, the data upload timestamp field must be within a preset recent time window of 7 days (168 hours) prior to the current moment. The submodule executes this query operation, performing conditional matching retrieval on the upload records of tens of thousands of display cabinets across the entire network, ignoring inactive or mismatched display cabinets, and locking all records that meet the above conditions. From the result set returned by the database, the submodule extracts the historical operating status parameters (including historical startup energy and voltage data) and the unique identification index information of each record, aggregating these discrete records into a temporary in-memory dataset to generate a dataset of active display cabinets of the same type. Table 1 shows some examples of the retrieved active data. Table 1: Results of Data Retrieval for Active Display Cases Record Number Display case model Identity Index Collection time Historical voltage RMS value Historical startup energy value 001 Model-850L DEV-001 2023-10-27-08:00 220.5 3.12 002 Model-850L DEV-002 2023-10-27-08:05 210.2 3.45 003 Model-850L DEV-003 2023-10-27-08:10 235.1 2.98 As shown in Table 1, the submodule retrieved multiple records of the same model of display cabinet, which included their voltage and energy data at different times.

[0026] The voltage window setting submodule calls the grid voltage effective value in the startup feature data pair as the reference point, performs addition and subtraction operations according to the preset voltage fluctuation tolerance parameters, calculates the reference voltage rise and fall limit values ​​respectively, and generates the same voltage screening deviation range. Based on the multi-source heterogeneous data environment gathered in the cloud, the submodule reads the real-time grid voltage effective value measured by the startup feature data pair, for example, a value of 215 volts, and sets it as the central reference point; the submodule reads the voltage fluctuation tolerance parameter preset in the system configuration file, which is set to 5 volts (or the specific voltage value converted in percentage form); the submodule performs an addition operation, adding the reference point of 215 volts to the tolerance of 5 volts, and obtains 220 volts as the upper limit; at the same time, it performs a subtraction operation, subtracting the reference point of 215 volts from the tolerance of 5 volts, and obtains 210 volts as the lower limit; the submodule combines these two boundary values ​​to construct a closed numerical range [210, 220], which defines the voltage range that is considered to be consistent with the current operating condition in terms of electrical characteristics, and generates the same voltage screening deviation range.

[0027] The working condition alignment and clustering submodule calls the same pressure screening deviation range, extracts the voltage value corresponding to the record and performs numerical inclusion relationship determination with the same pressure screening deviation range, retains the specific record whose voltage value falls within the range, separates the corresponding start-up energy value from the specific record and aggregates it into a single-dimensional numerical set to establish the same pressure and same type display cabinet data cluster. To achieve precise alignment of cloud-based big data with operating conditions, the submodule loads the previously generated dataset of active display cabinets of the same type, as well as the calculated same-voltage screening deviation intervals of 210 and 220. A traversal loop is initiated, sequentially reading each display cabinet record in the dataset and extracting the "historical effective voltage value" field from each record. The submodule then performs a numerical inclusion relationship determination between the extracted voltage value and the same-voltage screening deviation interval, checking if the voltage value is greater than or equal to 210 and less than or equal to 220. If the determination result is true, the record is marked as a "same-voltage sample," and the corresponding "historical startup energy value" is extracted. If the determination result is false, the record is skipped. After the traversal is complete, the submodule aggregates the startup energy values ​​of all marked records, forming a single-dimensional array set containing only numerical values. This set eliminates interference from voltage fluctuations and purely reflects the group startup characteristics of display cabinets of the same type under this specific voltage condition, establishing a data cluster of same-voltage, same-type display cabinets. For example, for the data in Table 1, the voltage of record 002, 210.2 volts, falls within the range, and its energy of 3.45 is retained; while the voltage of record 003, 235.1 volts, exceeds the range, and its energy of 2.98 is discarded.

[0028] Specifically, such as Figure 2 , 6 As shown, the population baseline generation module includes: The numerical sorting and organization submodule calls the data cluster of display cabinets of the same pressure and type, traverses the starting energy values ​​of the independent display cabinets included, performs numerical comparison operation, rearranges the energy data according to monotonically increasing logic, establishes the correspondence between numerical value and sequence position, removes the interference of time dimension index in the original data, and generates an ascending list of energy values. Leveraging the high concurrency processing capabilities of the cloud platform, a data cluster of similar display cabinets containing N startup energy values ​​is loaded into memory. For example, this cluster contains 500 specific energy values. The submodule uses the QuickSort algorithm to process these 500 values. The algorithm uses the first element of the array as the pivot, moving all values ​​smaller than the pivot to the left and values ​​larger than the pivot to the right, and recursively executing this process until the entire array is completely ordered. After performing the value comparison operation, the submodule rearranges the energy data according to a monotonically increasing logic from smallest to largest, generating a new ordered index list. In this list, index 0 corresponds to the smallest startup energy value, and index 499 corresponds to the largest startup energy value, thus establishing a strict correspondence between the value size and the sequence position. During this process, the submodule only retains the energy values ​​themselves, discarding the time dimension index and display cabinet ID information associated with the original data, eliminating interference from irrelevant variables, and generating an ascending list of energy values. For example, the first 5 digits of the sorted list are [2.85, 2.88, 2.90, 2.91, 2.92], and the last 5 digits are [4.10, 4.15, 4.20, 4.25, 4.50].

[0029] The distribution feature calculation submodule calls the ascending energy value list, locates the value point in the middle position according to the total length of the ascending energy value list as the central trend characterization, locates the value points in the quarter position and three-quarter position of the ascending energy value list respectively, performs a subtraction operation on the two quantile values ​​to obtain the difference reflecting the data dispersion width, combines the value at the central position and the dispersion width difference to form mathematical parameters for quantifying the population distribution state, and generates a distribution statistical feature set; The submodule mines group behavior characteristics and obtains the total length of the ascending energy value list, for example, length N=500. The submodule first calculates the center position index, i.e., 500 × 0.5 = 250, directly locating the value corresponding to the 250th position in the list. Assuming this value is 3.20, it is used as the median indicator Q2 to characterize the group's central tendency. Subsequently, the submodule calculates the quarter-position index (500 × 0.25 = 125) and the three-quarter-position index (500 × 0.75 = 375), locating the 125th position in the list... The submodule calculates the difference between the 375th and 375th quantile values ​​(assuming 3.05, or Q1) and the 375th quantile value (assuming 3.45, or Q3). It performs a subtraction operation on these two quantile values, subtracting the Q1 value from the Q3 value (3.45) to obtain a difference of 0.40. This difference is the interquartile range (IQR), used to obtain the difference reflecting the data dispersion width. The submodule combines and encapsulates the median index (3.20) and the interquartile range (0.40) to construct a set of mathematical parameters quantifying the population distribution state, generating a distribution statistical feature set.

[0030] The benchmark model construction submodule is based on the distribution statistical feature set. It maps the median value to the standard mechanical impedance reference line of the model display cabinet under a specific voltage, maps the interquartile range to the allowable reasonable fluctuation range width, and combines real-time sample capacity information to construct a multi-dimensional benchmark data structure to generate a group impedance distribution benchmark. The generation of a global baseline capable of being distributed or used for cloud-based diagnostics follows these steps: The submodule reads the median value (3.20) and interquartile range (0.40) from the distribution statistical feature set. It then maps the median value (3.20) directly to the standard mechanical impedance reference line for this model of display cabinet under approximately 215 volts, representing the typical startup energy consumption level of the display cabinet in a healthy state. The submodule also maps the interquartile range (0.40) to the allowable reasonable fluctuation range width, serving as a benchmark for determining the significance of individual differences. Furthermore, the submodule combines the current real-time sample size information (N=500) to weight the confidence level of the baseline, constructing a multi-dimensional data structure containing "baseline value = 3.20", "discrete scale = 0.40", and "sample size = 500". This structure defines the numerical profile of normal group operation, generating a group impedance distribution baseline.

[0031] Specifically, such as Figure 2 , 7 As shown, the wear outlier detection module includes: The impedance difference quantification submodule uses the median value index stored in the group impedance distribution benchmark as a reference anchor point, and combines the startup charge accumulation value recorded in the startup characteristic data with the median value index and the startup charge accumulation value to perform subtraction algebra operation on the median value index and the startup charge accumulation value to calculate the numerical difference between the real-time energy consumption of the target display cabinet in the startup event and the central trend of the same type of display cabinet group, and generates the energy benchmark deviation value. Combining cloud-based benchmarks and edge-uploaded data, the median value of 3.20 is extracted from the group impedance distribution benchmark as a reference anchor point. Simultaneously, the actual measured cumulative start-up charge value of the target display cabinet is retrieved from the start-up characteristic data pair, for example, a value of 3.90. The submodule performs a subtraction algebra operation on these two values, that is, subtracting the benchmark value of 3.20 from the target value of 3.90. Through this calculation, the submodule obtains a numerical difference of 0.70. This difference directly quantifies the absolute distance between the actual energy consumed by the target display cabinet in this start-up event and the group center trend of similar display cabinets under the same voltage conditions. A positive value indicates that the energy consumption is too high, and a negative value indicates that it is too low, generating an energy benchmark deviation value.

[0032] The outlier normalization submodule uses the energy baseline deviation value as a benchmark to call the interquartile range index recorded in the population impedance distribution benchmark as a measure of the data dispersion. It performs a division operation with the energy baseline deviation value as the numerator and the interquartile range index as the denominator, transforming the absolute difference in the physical energy dimension into the relative distance in the statistical dimension, and generating the impedance deviation coefficient. Statistical tools are used to eliminate individual differences, obtaining the previously calculated energy baseline deviation value of 0.70. The interquartile range (IQR) index of 0.40 recorded in the group impedance distribution baseline is then used as a scale to measure the degree of data dispersion. The submodule performs a division operation, using the energy baseline deviation value of 0.70 as the numerator and the interquartile range index of 0.40 as the denominator, i.e., calculating 0.70 / 0.40. This operation converts the energy difference with physical units into a dimensionless statistical multiple, with a result of 1.75. This value indicates that the deviation of the target display cabinet is 1.75 times the normal fluctuation range of the group. Through this normalization process, the differences in energy baseline under different models or voltage levels are eliminated, achieving a unified standard for outlier measurement and generating an impedance deviation coefficient.

[0033] The wear status determination submodule uses the impedance deviation coefficient to read the preset mechanical wear anomaly determination threshold. It compares the impedance deviation coefficient with the anomaly determination threshold. If the impedance deviation coefficient is greater than the anomaly determination threshold, the anomaly marking logic is triggered. The corresponding risk severity level is matched according to the magnitude of the impedance deviation coefficient exceeding the anomaly determination threshold, and the wear warning determination result of the display cabinet is generated. The final decision in the cloud-edge collaborative closed loop is completed. The calculated impedance deviation coefficient of 1.75 is read, and the preset mechanical wear anomaly judgment thresholds are read from the system's fault diagnosis strategy library. For example, the warning threshold is set to 1.5 and the alarm threshold is set to 3.0. The submodule performs a numerical comparison operation between the impedance deviation coefficient of 1.75 and these anomaly judgment thresholds. First, 1.75 is compared with 1.5. It is determined that 1.75 is greater than 1.5, triggering the anomaly marking logic. Then, 1.75 is compared with 3.0. It is determined that 1.75 is less than 3.0. According to the extent that the impedance deviation coefficient exceeds the anomaly judgment threshold, the submodule matches the corresponding risk severity level as "Level 1 Slight Wear Warning". The submodule packages this level information together with the deviation coefficient and the reference benchmark to construct a structured message containing the diagnostic conclusion and generate the display cabinet wear warning judgment result. Table 2 shows the numerical example of the judgment logic of this module. Table 2: Logic Example Table for Wear Determination Parameters Values / Settings illustrate Target starting energy 3.90 Measured value Group baseline median 3.20 Reference Anchor Point Population interquartile range (IQR) 0.40 Discrete scale Energy reference deviation value 0.70 Calculate: 3.90 − 3.20 Impedance deviation coefficient 1.75 Calculate: 0.70 / 0.40 Warning threshold 1.5 Setting value alarm threshold 3.0 Setting value Judgment Result Level 1 minor wear 1.5<1.75<3.0 As shown in Table 2, the experimental results show that by converting physical deviations into relative coefficients in units of population dispersion, it is possible to accurately locate display cases in a sub-healthy state.

[0034] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of protection of the technical solution.

Claims

1. A remote intelligent operation and maintenance management system for commercial display cabinets based on big data, characterized in that: The system includes: The edge waveform acquisition module monitors the closing action status of the relay control signal of the terminal compressor of the commercial display cabinet, acquires transient current and real-time voltage waveform data, and generates a startup transient waveform data package; The startup energy calculation module performs integration on the transient current waveform data packet to obtain the startup energy, performs root mean square calculation on the real-time voltage waveform data to obtain the voltage RMS value, and associates the startup energy and voltage RMS value to generate startup feature data pairs. The same-pressure working condition clustering module parses the model identifier, retrieves active data with the model identifier, sets the voltage deviation range based on the effective voltage value, filters active data that meet the voltage deviation range to extract the start-up energy, and establishes a data cluster of display cabinets of the same type and under the same pressure. The group baseline generation module sorts the starting energy in the data clusters of the same type of display cabinets under the same pressure, calculates the median index and interquartile range index, and constructs a group impedance distribution benchmark. The wear outlier determination module calculates the difference between the starting energy and the median index in the group impedance distribution benchmark, divides the median index difference by the interquartile range index to obtain the deviation index, compares the deviation index with the preset abnormal threshold, and generates a wear warning determination result for the display cabinet.

2. The remote intelligent operation and maintenance management system for commercial display cabinets based on big data as described in claim 1, characterized in that, The startup transient waveform data packet includes a collection trigger timestamp, a dual-channel sampling point array, and sensor calibration parameters. The startup characteristic data pair includes mechanical work characterization values, grid power supply capacity characterization values, and time-series correlation tags. The same-voltage and same-type display cabinet data cluster includes a display cabinet identification set of the same model, an approximate voltage operating condition record set, and a startup energy value sequence. The group impedance distribution benchmark includes the group energy median line, impedance dispersion range, and statistical sample capacity identifier. The display cabinet wear warning judgment result includes the mechanical impedance abnormality level, display cabinet maintenance suggestion code, and fault occurrence probability assessment value.

3. The remote intelligent operation and maintenance management system for commercial display cabinets based on big data as described in claim 1, characterized in that, The edge waveform acquisition module includes: The trigger signal capture submodule monitors the closing action status of the compressor relay control signal of the commercial display cabinet terminal. When the rising state of the signal representing the closing action is detected, the clock is locked, real-time time point data is obtained as a reference, and the time point data is logically associated with the hardware start request signal to generate a data acquisition synchronization trigger command. The dual-channel synchronous acquisition submodule responds to the acquisition synchronization trigger command, synchronously activates the current transformer and voltage sensor connected to the compressor power supply line, continuously reads the analog signal within a fixed time window during the startup process according to the preset sampling frequency, converts the analog signal into digital quantized data through the analog-to-digital converter and establishes the corresponding time index, and generates an electrical transient waveform sequence. The data packet encapsulation submodule calls the electrical transient waveform sequence, performs time-domain synchronization alignment operation on the current change trajectory and voltage fluctuation trajectory according to the time index, extracts the sampling frequency and display cabinet identifier of the waveform sequence as header metadata, and integrates the aligned dual-channel waveform data with the header metadata through binary serialization to construct the startup transient waveform data packet.

4. The remote intelligent operation and maintenance management system for commercial display cabinets based on big data as described in claim 3, characterized in that, The startup energy calculation module includes: The current integration operation submodule calls the startup transient waveform data packet, deconstructs the transient current waveform data sequence and timestamp index, performs a product operation on the instantaneous current amplitude of each discrete sampling point and the time interval between adjacent sampling points, performs an accumulation and summation operation on the product terms within the preset startup time window, quantifies the total integral of current with time during the process of the compressor rotor accelerating from a stationary state to the rated speed, and generates the startup charge accumulation value. The voltage root mean square calculation submodule responds to the calculation completion signal of the accumulated starting charge value, extracts real-time grid voltage waveform data synchronized with the energy value time span, performs a square operation on each discrete voltage sampling point in the real-time grid voltage waveform data sequence, calculates the arithmetic mean of the square values ​​and performs square root processing on the mean, obtains the voltage physical quantity characterizing the real-time load capacity of the power supply network at the moment of startup, and generates the effective value of grid voltage. The feature association packaging submodule establishes a logical mapping relationship between the two values ​​based on the effective value of the grid voltage and the unique identifier of the display cabinet startup event. It marks the startup charge accumulation value as a mechanical load characteristic index and the effective value of the grid voltage as an input voltage condition index. According to the preset communication protocol format, the indexes are combined into a structured data unit with spatiotemporal correlation to generate startup feature data pairs.

5. The remote intelligent operation and maintenance management system for commercial display cabinets based on big data as described in claim 4, characterized in that, The same pressure condition clustering module includes: The same type data retrieval submodule calls the startup feature data pair, parses the target display cabinet model identifier information, performs condition matching retrieval for display cabinet records across the entire network, locks records with the same model identifier and valid upload behavior within a preset recent time window, extracts the corresponding running status parameters and identity index information of the records, and generates a dataset of active display cabinets of the same type. The voltage window setting submodule calls the grid voltage effective value in the startup feature data pair as the reference point, performs addition and subtraction operations according to the preset voltage fluctuation tolerance parameter, calculates the reference voltage rise and fall limit values ​​respectively, and generates the same voltage screening deviation range. The working condition alignment clustering submodule calls the same pressure screening deviation range, extracts the voltage value corresponding to the record and performs numerical inclusion relationship determination with the same pressure screening deviation range, retains specific records whose voltage values ​​fall within the range, separates the corresponding start-up energy values ​​from the specific records and aggregates them into a single-dimensional numerical set to establish a same pressure and same type display cabinet data cluster.

6. The remote intelligent operation and maintenance management system for commercial display cabinets based on big data as described in claim 5, characterized in that, An inverted index structure is established for display cabinet records across the entire network. The inverted index structure is used to locate candidate data whose timestamp attributes fall within a continuous time interval. The system checks whether the candidate data includes a complete starting current waveform file and the corresponding compressor operating impedance data, and only records containing complete data are marked as having valid upload behavior.

7. The remote intelligent operation and maintenance management system for commercial display cabinets based on big data as described in claim 5, characterized in that, The population baseline generation module includes: The numerical sorting and sorting submodule calls the data cluster of the same pressure and type of display cabinet, traverses the starting energy values ​​of the independent display cabinets, performs a numerical comparison operation, rearranges the energy data according to monotonically increasing logic, establishes the correspondence between numerical value and sequence position, removes the time dimension index interference in the original data, and generates an ascending energy value list. The distribution feature calculation submodule calls the ascending energy value list, locates the value point in the middle position according to the total length of the ascending energy value list as the central trend characterization, locates the value points in the quarter position and three-quarter position of the ascending energy value list respectively, performs subtraction operation on the two quantile values ​​to obtain the difference reflecting the data dispersion width, combines the central position value and the dispersion width difference to form mathematical parameters for quantifying the population distribution state, and generates a distribution statistical feature set; The benchmark model construction submodule, based on the aforementioned distribution statistical feature set, maps the median value to the standard mechanical impedance reference line of the model display cabinet under a specific voltage, maps the interquartile range to the allowable reasonable fluctuation range width, and constructs a multi-dimensional benchmark data structure by combining real-time sample capacity information to generate a group impedance distribution benchmark.

8. The remote intelligent operation and maintenance management system for commercial display cabinets based on big data as described in claim 7, characterized in that, The process of constructing a multi-dimensional benchmark data structure by combining real-time sample size information is as follows: call the logarithmic calculation logic to calculate the logarithm of the list length value with base ten, and multiply the logarithm by the preset confidence weight coefficient to obtain the sample confidence score. A structured storage space is allocated in memory, and a standard mechanical impedance reference line, the allowable reasonable fluctuation range width, and the sample confidence score are written in sequence. The minimum and maximum values ​​in the ascending energy value list are written into the structured storage space as boundary constraint parameters to form the group impedance distribution benchmark including five-dimensional data fields.

9. The remote intelligent operation and maintenance management system for commercial display cabinets based on big data as described in claim 7, characterized in that, The wear outlier detection module includes: The impedance difference quantification submodule uses the median value index stored in the group impedance distribution benchmark as a reference anchor point, and combines the startup charge accumulation value recorded in the startup characteristic data to perform a subtraction algebra operation on the median value index and the startup charge accumulation value to calculate the numerical difference between the real-time energy consumption of the target display cabinet in the startup event and the central trend of the same type of display cabinet group, and generates an energy benchmark deviation value. The outlier normalization submodule, based on the energy reference deviation value, calls the interquartile range index recorded in the population impedance distribution reference as a scale to measure the degree of data dispersion. It performs a division operation with the energy reference deviation value as the numerator and the interquartile range index as the denominator, transforming the absolute difference in the physical energy dimension into a relative distance in the statistical dimension, and generating an impedance deviation coefficient. The wear status determination submodule uses the impedance deviation coefficient to read the preset mechanical wear anomaly determination threshold, performs a numerical comparison operation between the impedance deviation coefficient and the anomaly determination threshold, and triggers the anomaly marking logic if the impedance deviation coefficient is greater than the anomaly determination threshold. Based on the magnitude of the impedance deviation coefficient exceeding the anomaly determination threshold, the corresponding risk severity level is matched, and a wear warning determination result for the display cabinet is generated.

10. The remote intelligent operation and maintenance management system for commercial display cabinets based on big data as described in claim 9, characterized in that, The process of performing subtraction algebraic operation on the median index and the cumulative value of the starting charge is as follows: read the cumulative value of the starting charge recorded in the starting characteristic data as the minuend, read the median index stored in the group impedance distribution reference as the subtrahend, and perform double-precision floating-point subtraction to obtain the original difference value. A unidirectional wear drift filtering logic is constructed to determine whether the original difference value is greater than 0. When the original difference value is greater than 0, it indicates that there is an increase in positive impedance in the display cabinet, and the original difference value is assigned as the energy reference deviation value. When the original difference value is less than or equal to 0, it indicates that the display case is in a low impedance operating state. The energy reference deviation value is forcibly set to 0 to eliminate negative deviations caused by non-wear factors.

Images