A centralized control method and system for hot gas defrosting

Abstract

This invention discloses a centralized control method and system for hot gas defrosting. Through data acquisition, frost layer prediction and triggering, defrosting execution, and optimization feedback steps, it achieves intelligent and precise control of the defrosting process. The system includes a sensor group, an edge data acquisition module, a local controller, a communication module, a cloud-based centralized control platform, and actuators. This invention employs a fusion prediction model, combining rule-based rapid scoring with machine learning-based accurate prediction to dynamically assess frost risk and trigger defrosting, avoiding the blindness of traditional timed defrosting. Simultaneously, it ensures system stability through grouped sequential control and safety monitoring, and utilizes a cloud platform to aggregate multi-site data for global optimization scheduling and model iteration. This invention effectively reduces defrosting energy consumption, maintains stable storage / cabinet temperatures, and improves the operating efficiency and intelligent management level of the refrigeration system.

Description

Technical Field

[0001] This invention relates to the field of refrigeration system control technology, and in particular to a centralized control method and system for hot gas defrosting. Background Technology

[0002] Currently, vapor compression refrigeration systems are widely used in commercial refrigeration (such as supermarket display cases and cold storage). As the main heat exchange component of the system, the evaporator inevitably absorbs moisture from the air and condenses it into frost on its fins when operating at low temperatures. The accumulation of frost significantly increases thermal resistance, obstructs airflow, leads to decreased refrigeration efficiency and increased energy consumption, and causes adverse temperature fluctuations inside the cabinet or storage unit, affecting the quality and safety of stored goods.

[0003] However, existing technologies have the following drawbacks: Inaccurate defrosting timing: Traditional defrosting relies heavily on fixed timing or manual experience, failing to consider real-time factors such as actual frost thickness, door opening and closing frequency, ambient temperature and humidity changes, and cargo load fluctuations. This results in a large amount of "ineffective defrosting," leading to energy waste; or "incomplete defrosting," causing decreased refrigeration efficiency, large fluctuations in warehouse / cabinet temperature, and increased risk of food spoilage.

[0004] Poor system coordination: Although hot gas defrosting utilizes the waste heat of the compressor exhaust and is more energy-efficient than electric heating, in systems with multiple evaporators or multiple cold rooms, if defrosting is carried out simultaneously, it can easily cause problems such as drastic fluctuations in system pressure, compressor liquid slugging, and poor oil return, making it difficult to guarantee the safety and operational stability of the system.

[0005] Fragmented data acquisition: Existing systems mostly monitor only a few temperature or pressure points, lacking comprehensive monitoring of key parameters such as energy consumption, evaporator differential pressure, ambient humidity, and number of door openings. This prevents the system from performing predictive maintenance and in-depth energy consumption optimization.

[0006] Decentralized management: Each refrigeration unit is usually managed by an independent PLC or controller, making it difficult to achieve unified scheduling, remote monitoring, peak-valley electricity pricing strategy optimization, and centralized fault early warning for the entire chain of supermarkets or multiple cold storage facilities.

[0007] In view of this, the present invention is proposed. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a centralized control method for hot gas defrosting, comprising the following steps: To achieve the above objectives, the technical solution of the present invention is as follows: Step 1, Data Acquisition Step: Real-time acquisition of multi-dimensional operating parameters of the refrigeration system, including at least the evaporator inlet and outlet temperature difference, evaporator differential pressure, ambient relative humidity, opening status of the refrigeration unit door or cabinet door, evaporator surface temperature, compressor exhaust parameters, and system energy consumption. Step 2, Frost Prediction and Triggering Step: Based on the collected multi-dimensional operating parameters, the fusion prediction model is used to dynamically assess the frost risk of each evaporator. When the frost risk reaches the preset triggering condition, a defrosting command is generated. The fusion prediction model includes a rule-based fast scoring model and / or a machine learning-based accurate prediction model. Step 3, Defrosting Execution Steps: In response to the defrosting command, control the target evaporator to execute a hot gas defrosting sequence, which includes: shutting off the liquid supply, stopping or reducing the fan speed, opening the hot gas bypass valve, monitoring the defrosting process parameters, and closing the hot gas valve and restoring cooling when the termination conditions are met. Step 4: Optimize the feedback process: Record the actual time and energy consumption of the defrosting process, and upload it as feedback data to the cloud-based centralized control platform for updating and optimizing the fusion prediction model.

[0009] Preferably, the rule-based fast scoring model specifically comprises: Based on multiple preset risk factors and their weights, the current real-time data is weighted and scored to obtain a risk score. The risk factors include whether the temperature difference between the inlet and outlet of the evaporator exceeds the first temperature difference threshold, whether the differential pressure across the evaporator exceeds the first differential pressure threshold, whether the relative humidity of the environment exceeds the first humidity threshold, whether the time since the end of the last defrost exceeds the first time threshold, and whether the cumulative time of door opening within the preset period exceeds the first time threshold. The risk score is compared with a preset risk level threshold. If it exceeds the high risk threshold, it is determined to be high risk, and a defrosting command is immediately triggered. If it exceeds the medium risk threshold but does not reach the high risk threshold, it is determined to be medium risk, and the frequency of data collection and monitoring is increased.

[0010] Preferably, the machine learning-based accurate prediction model is a Long Short-Term Memory (LSTM) network model or a Random Forest (RF) model; the LSTM model takes historical time series data as input and outputs a predicted value for the future frost growth rate or frost probability; the RF model takes the multidimensional features and statistical lag features of the current moment as input and outputs the frost melting probability; the triggering condition is that the predicted frost growth rate or frost probability exceeds a preset machine learning triggering threshold.

[0011] Preferably, the defrosting execution step further includes a grouping sequence control sub-step: when multiple evaporators need to be defrosted at the same time, the evaporators to be defrosted are divided into multiple defrosting groups, and only one group or a preset number of defrosting groups are allowed to execute the hot gas defrosting sequence at the same time, while the remaining defrosting groups are in a waiting state, so as to avoid excessive compressor load and drastic fluctuations in system pressure.

[0012] Preferably, the defrosting execution step further includes a safety monitoring sub-step: During hot gas defrosting, the evaporator surface temperature, system exhaust pressure, and suction superheat are monitored in real time. When the evaporator surface temperature rises to the preset defrosting completion temperature threshold, the defrosting is considered to be over. When the exhaust pressure exceeds the safety limit or the suction superheat is lower than the liquid slugging risk threshold, the current defrosting process is immediately interrupted and an alarm signal is issued.

[0013] Preferably, the optimization feedback step further includes a cloud-based centralized optimization sub-step: the cloud-based centralized control platform aggregates data uploaded from multiple refrigeration stations, combines time-of-use electricity price information and frost risk prediction for each station, uses an optimization algorithm to generate a global defrosting scheduling plan, and sends the adjusted control parameters to the local controller.

[0014] A centralized hot gas defrosting control system, used to implement the method according to any one of claims 1 to 6, characterized in that it comprises: Sensor arrays are deployed at various monitoring points in the refrigeration system to collect multi-dimensional operating parameters in real time. An edge data acquisition module, connected to the sensor group, is used for preprocessing and feature extraction of the acquired data; A local controller, connected to the edge data acquisition module, is used to execute local control logic, including receiving defrosting commands and controlling the actuator to complete the hot gas defrosting sequence; The communication module is used to enable data exchange between the local controller and the cloud-based centralized control platform; A cloud-based centralized control platform, connected to the communication module, is used to store and analyze all uploaded data, run the fusion prediction model, generate and issue defrosting commands, and perform global optimization scheduling. The actuator, connected to the local controller, includes at least a hot gas solenoid valve, a liquid supply solenoid valve, an electronic expansion valve, and a fan, and is used to perform corresponding switching or regulating actions in response to control signals.

[0015] The sensor group includes at least: a temperature sensor for measuring the inlet and outlet temperatures of the evaporator, a differential pressure transmitter for measuring the differential pressure across the evaporator, a humidity sensor for measuring the relative humidity of the environment, a door magnetic switch or infrared sensor for counting the opening status of the storage door or cabinet door, a temperature sensor for measuring the surface temperature of the evaporator, a pressure transmitter and a temperature sensor for measuring the compressor discharge pressure and temperature, and an energy meter for measuring the system energy consumption.

[0016] The local controller is a programmable logic controller (PLC) or an embedded controller with edge computing capabilities. It can independently run the rule-based fast scoring model and maintain local automatic control when communication with the cloud is interrupted.

[0017] The cloud-based centralized control platform also integrates a data visualization module and a predictive maintenance module. The data visualization module provides real-time / historical data queries, energy consumption reports, and a multi-site group management interface. The predictive maintenance module assesses the health status of equipment and issues maintenance warnings based on long-term data analysis.

[0018] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art: 1. This invention uses intelligent on-demand defrosting instead of traditional timed defrosting, which can reduce the number of ineffective defrostings by more than 30%, significantly reducing the additional heat load and fan energy consumption caused by defrosting. At the same time, grouped sequential control avoids drastic fluctuations in compressor load, effectively reducing peak system power consumption and overall operating costs.

[0019] 2. This invention accurately predicts the risk of frost formation and addresses it before the frost layer affects refrigeration efficiency, preventing heat transfer deterioration and significant fluctuations in storage / cabinet temperature due to excessive frost thickness. A rigorous safety monitoring mechanism effectively prevents liquid slugging and high-pressure malfunctions, ensuring food storage safety and stable equipment operation. 3. This invention integrates a predictive model that continuously optimizes itself using cloud-based big data, resulting in increasingly precise control strategies. The system supports centralized monitoring of multiple sites, remote scheduling, and predictive maintenance, significantly reducing manual inspection costs and substantially improving the intelligent operation and maintenance level of the refrigeration system. Attached Figure Description

[0020] The accompanying drawings, as part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention, but do not constitute an undue limitation of the invention. Obviously, the drawings described below are merely some embodiments, and those skilled in the art can obtain other drawings based on these drawings without creative effort. In the drawings: Figure 1 This is a schematic diagram of the process and structure of the present invention; Figure 2 This is the main flowchart of the centralized control method for hot gas defrosting of the present invention; Figure 3 This is a flowchart of the hot gas defrosting execution sequence of the present invention; Figure 4 A detailed flowchart of the rule-based fast scoring model of the present invention. Figure 5 This is a flowchart of the cloud-based centralized optimization and feedback process of the present invention; Figure 6 This is a hardware system block diagram of the present invention.

[0021] Explanation of reference numerals in the attached figures: 1. Sensor group; 2. Edge data acquisition module; 3. Local controller; 4. Communication module; 5. Cloud-based centralized control platform; 6. Actuator.

[0022] It should be noted that these accompanying drawings and textual descriptions are not intended to limit the scope of the invention in any way, but rather to illustrate the concept of the invention to those skilled in the art by referring to specific embodiments. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments will be clearly and completely described below with reference to the accompanying drawings. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0024] like Figure 6 As shown, this embodiment provides a centralized control method and system for hot gas defrosting, wherein the centralized control system for hot gas defrosting includes the following components: Sensor group 1: Deployed at various monitoring points in the refrigeration system to collect multi-dimensional operating parameters in real time; specifically includes: temperature sensors (such as PT100 or thermocouples); installed at the evaporator inlet, outlet, evaporator coil surface, compressor exhaust port, compressor suction port, etc., to measure the evaporator inlet and outlet temperatures, evaporator surface temperature, exhaust temperature, and suction temperature.

[0025] Differential pressure transmitter: installed before and after the evaporator, used to measure the pressure difference ΔP between the inlet and outlet of the evaporator.

[0026] Humidity sensor: Installed in the external environment and inside the cold storage or freezer to measure the relative humidity (RH).

[0027] Door status sensor: Door magnetic switch or infrared beam sensor, used to count the number of times the warehouse door or cabinet door is opened and the cumulative opening time.

[0028] Pressure transmitter: Installed at the compressor's exhaust port and suction port, used to measure exhaust pressure and suction pressure.

[0029] Energy meter: Installed on the main power supply line of the refrigeration system to measure the real-time energy consumption of the system.

[0030] Edge data acquisition module 2: Connected to sensor group 1, it is responsible for preprocessing the acquired raw data, including filtering (such as moving average filtering), outlier removal, normalization, and feature extraction (such as calculating temperature difference ΔT, pressure difference ΔP, and cumulative door opening time). The edge module supports industrial bus protocols such as Modbus RTU / TCP and CAN, and can acquire data at a high sampling frequency (such as 1 second).

[0031] Local Controller 3: Can be a Programmable Logic Controller (PLC) or an embedded industrial controller. Local Controller 3 is integrated with or separate from Edge Data Acquisition Module 2, and is used to execute local control logic, including receiving defrosting commands and controlling the actuators to complete the hot gas defrosting sequence; the local controller has edge computing capabilities, can independently run a rule-based fast scoring model, and maintain local automatic control when communication with the cloud is interrupted, ensuring system reliability.

[0032] Communication module 4: Used to realize data exchange between the local controller 3 and the cloud-based centralized control platform 5. Communication methods include, but are not limited to, 4G / 5G, Wi-Fi, Ethernet, LoRa, NB-IoT, etc. The communication module is responsible for periodically (e.g., every minute) uploading the preprocessed feature data to the cloud, and receiving control parameters, instructions, and model updates from the cloud.

[0033] Cloud-based centralized control platform 5: A server cluster deployed in a public or private cloud, including real-time databases, historical databases, model training servers, and application servers. The cloud platform is responsible for: Store and analyze all uploaded data. Run fusion prediction models (including machine learning models) and generate defrosting commands for issuance; perform global optimization scheduling (such as generating defrosting plans by combining time-of-use pricing); provide a data visualization module (Web / App interface) and a predictive maintenance module.

[0034] Actuator 6: Connected to the local controller 3, it performs corresponding switching or regulating actions in response to control signals. It mainly includes: a hot gas solenoid valve: controlling the opening and closing of the hot gas bypass line; a refrigerant supply solenoid valve: controlling the opening and closing of the refrigerant supply line; an electronic expansion valve (EEV): precisely regulating the refrigerant flow rate; an evaporator fan: speed adjustable, used to control the airflow; and a pressure relief solenoid valve: used to release residual high-pressure gas in the evaporator after defrosting.

[0035] like Figures 1 to 5 As shown, the centralized control method for hot gas defrosting of the present invention includes the following steps: Step 1: Data acquisition and feature extraction: Real-time acquisition of multi-dimensional operating parameters of the refrigeration system, including at least: The temperature difference between the inlet and outlet of the evaporator is ΔT = T_in - T_out (in the cooling mode, T_in > T_out, and ΔT is a positive value; as the frost layer thickens, the heat exchange efficiency decreases, and ΔT will gradually decrease, so the trend of ΔT can be used to characterize the degree of frost). The differential pressure ΔP (Pa) across the evaporator increases as the frost layer thickens, leading to increased airflow resistance and a rise in ΔP; the relative humidity RH (%); the opening status of the storage door or cabinet door, and the cumulative door opening time within a preset time period (e.g., the past hour) door_1h; and the evaporator surface temperature T_surf. Compressor discharge pressure P_discharge, discharge temperature T_discharge, suction pressure P_suction, suction temperature T_suction; system energy consumption E (kWh).

[0036] The edge data acquisition module preprocesses the data (filtering, normalization) and calculates derived features, such as the time since the last defrost (t_since), compressor load rate (current capacity / total installed capacity), and suction superheat (SH=T_suction-T_sat(P_suction)). The preprocessed data is then uploaded to the cloud via the communication module.

[0037] Step 2: Frost Prediction and Triggering Decision: Based on the collected multi-dimensional operating parameters, the fusion prediction model is used to dynamically assess the frost risk of each evaporator. The fusion prediction model includes a rule-based fast scoring model and a machine learning-based accurate prediction model, which can work independently or in conjunction.

[0038] 2.1 Rule-based fast scoring model: This model runs on a local PLC for real-time and rapid risk assessment. The specific rules are as follows: based on multiple preset risk factors and their weights, the current real-time data is weighted and scored to obtain a risk score.

[0039]

[0040] in This indicates an indicator function (if it satisfies = 1, otherwise = 0); Risk factors include: Temperature difference factor: Whether ΔT is lower than the first temperature difference threshold (e.g., 2.0℃), weight w1=3.0; Pressure differential factor: Whether ΔP exceeds the first pressure differential threshold (e.g., 150 Pa), weight w2=3.5; Humidity factor: Whether the RH exceeds the first humidity threshold (e.g., 85%), weight w3=2.0; Time factor: Whether t_since exceeds the first time threshold (e.g., 240 minutes), weight w4=2.0; Door opening factor: Whether the cumulative door opening time (door_1h) within a preset time period (e.g., the past 1 hour) exceeds the first time threshold (e.g., 30 minutes), with a weight of w5=1.5.

[0041] The scoring formula is: Score = w1·I + w2·I + w3·I + w4·I + w5·I where...

[0042] Compare the risk score with a preset risk level threshold: If the score is ≥ 7.0 (high-risk threshold), it is considered high-risk and a defrost command is triggered immediately; if the score is 4.0 ≤ Score < 7.0 (medium-risk threshold), it is considered medium-risk and the data collection and monitoring frequency is increased (e.g., from once every 5 minutes to once every 1 minute). If it remains medium-risk for 30 minutes, defrost is triggered; if the score is < 4.0, it is considered low-risk and cooling continues.

[0043] A precise prediction model based on machine learning: This model mainly runs on a centralized cloud control platform and can be trained using historical data to improve prediction accuracy. The model can employ Long Short-Term Memory (LSTM) networks or Random Forest (RF). LSTM model: Suitable for time series prediction. The input is a multi-dimensional feature sequence (including ΔT, ΔP, RH, door_1m, ambient temperature T_amb, etc.) from the past N time steps (e.g., the past 6 hours, step size 10 minutes), and the output is the frost growth rate or frost probability P_frost for a future period (e.g., the next 1 hour). The model structure includes LSTM layers, fully connected layers, and a Sigmoid output layer. During training, historical running data and corresponding actual frost labels (obtainable through experimental calibration or image recognition) are used. When P_frost exceeds a preset machine learning trigger threshold (e.g., 0.7), defrosting suggestions can be generated. LSTM (Long Short-Term Memory Network): Processes time series data, taking the past N-step feature sequence (ΔT, ΔP, RH, door opening, etc.) as input, and predicting the future frost growth rate or risk probability. The model equation is simplified to:

[0044] in, For the eigenvector at time t, In hidden state, To predict the probability of risk, It is activated by sigmoid. It is suitable for capturing long-term dependencies such as peak opening times and environmental changes. The model is trained using historical running data and external factors (weather, peak and off-peak electricity prices), and thresholds / weights are periodically distributed to the edges.

[0045] The frost growth rate or frost probability is calculated: the frost thickness growth is modeled using differential equations, taking into account humidity, temperature difference, and door opening interference; , In the above formula: Frost layer thickness (mm) Time (h) mm / h / unit (growth coefficient, which can be adjusted based on actual measurements), Relative humidity (%) (Critical humidity threshold) (Evaporation temperature) (Ambient temperature) (Opening-up impact factor) Door opening percentage (0-1, e.g., 0.2 means the door is open 20% of the time), indirect parameter calculation: ; , , Pa, Pa / mm.

[0046] Random Forest (RF) Model: Suitable for non-time-series current-state classification / regression. Inputs are multi-dimensional features and statistical lag features at the current moment (such as averages and maximum values ​​over the past hour), and output is the defrost probability. RF models are robust and can run in the cloud or on edge controllers with strong computing power. Defrost is triggered when the output probability exceeds a threshold. Random Forest (RF): Multi-feature non-linear classification / regression, robust, and with good interpretability (feature importance analysis). Inputs are current + statistical lag features, and outputs the defrost probability. Tree ensemble formula:

[0047] Where B represents the number of trees, and represents a single decision tree. Model training uses historical operational data plus external factors (weather, peak and off-peak electricity prices), and thresholds / weights are periodically distributed to the edges.

[0048] The cloud platform can incrementally train the model using new data daily or weekly, update the model weights, and send the updated model parameters or trigger thresholds to the local controller.

[0049] Triggering Decision Fusion Logic: In practical applications, a strategy of "primarily local rapid scoring, supplemented by accurate cloud prediction" can be adopted. Local real-time triggering is based on rule-based scoring, while cloud prediction is used to optimize scheduling (such as recommending defrosting during periods of low electricity prices) and dynamically adjust local thresholds. When the local score reaches medium risk and the cloud prediction probability is high, triggering can also be initiated in advance.

[0050] Defrosting Execution: In response to the defrosting command, the local controller controls the target evaporator to execute a hot gas defrosting sequence, specifically including the following sub-steps: Grouping sequence control sub-step; when multiple evaporators need to defrost simultaneously, to avoid excessive compressor load and drastic system pressure fluctuations, the evaporators to be defrosted are divided into multiple defrosting groups (grouping can be based on refrigerant piping, physical location, or logical grouping). Only one group or a preset number (e.g., 1-2 groups) of defrosting groups are allowed to execute the hot gas defrosting sequence simultaneously, while the remaining defrosting groups are in a waiting state. The system checks the current compressor load rate; if the load rate is higher than a safety threshold (e.g., 90%), defrosting is delayed.

[0051] Defrosting sequence: Preparation before defrosting: Close the liquid supply solenoid valve of the target evaporator and set the opening of the electronic expansion valve (EEV) to 0; stop the evaporator fan or reduce its speed to a very low speed (such as 10%) to reduce cooling loss.

[0052] To activate the hot gas: Open the main hot gas solenoid valve (if not already open), then open the group hot gas solenoid valve of the target evaporator to allow the high-temperature, high-pressure refrigerant vapor discharged from the compressor to enter the evaporator and begin melting the frost layer.

[0053] Process monitoring: Real-time monitoring of evaporator surface temperature T_surf, system discharge pressure P_discharge, and suction superheat SH. When T_surf rises to the preset defrosting completion temperature threshold (e.g., 5℃~10℃), defrosting is considered complete; simultaneously, a maximum defrosting duration t_max (e.g., 20 minutes) is set to prevent excessive defrosting time. If the discharge pressure exceeds the safety limit (e.g., 1.8MPa) or the suction superheat falls below the liquid slugging risk threshold (e.g., 2K), the current defrosting process is immediately interrupted, an alarm signal is issued, and fault handling is initiated.

[0054] Defrosting End and Recovery: Close the group hot gas solenoid valve, open the pressure relief solenoid valve (delay for 30 seconds) to release residual high-pressure gas; close the pressure relief valve, reopen the liquid supply solenoid valve, and restore EEV regulation; after a 60-second delay, gradually restore the fan speed to normal.

[0055] Record feedback: Record the actual defrosting time t_defrost and the energy consumption ΔE during the defrosting process.

[0056] Step 4: Optimize Feedback: Upload the recorded defrosting time and energy consumption as feedback data to the cloud-based centralized control platform for updating and optimizing the fusion prediction model.

[0057] 4.1 Cloud-based centralized optimization sub-step: The cloud platform aggregates data uploaded from multiple refrigeration stations, combines time-of-use electricity price information (peak, flat, and valley periods) and frost risk prediction for each station, and uses optimization algorithms (such as genetic algorithms, particle swarm optimization, or reinforcement learning) to generate a global defrosting scheduling plan. The defrosting operation is scheduled during periods of low electricity prices and low ambient humidity as much as possible. The optimized control parameters (such as risk thresholds and triggering conditions) can be sent to the local controller for dynamic adjustment.

[0058] 4.2 Predictive maintenance and cloud platforms assess equipment health based on long-term data analysis. For example, monitoring the historical trend of ΔP for a single evaporator, if ΔP slowly increases by more than a certain percentage (e.g., 30%) over a continuous period, it is determined that the fan filter is clogged or the evaporator is dirty, generating a cleaning warning. If the ratio of compressor operating current to suction pressure deviates from the baseline curve, it can warn of refrigerant leakage or efficiency reduction. The predictive maintenance module pushes warning information to maintenance personnel through a visual interface (Web / App).

[0059] This invention achieves on-demand defrosting through a fusion predictive model, completely changing the traditional, inefficient timed defrosting method. It reduces ineffective defrosting by over 30%, significantly lowering the additional heat load and fan energy consumption caused by hot gas defrosting. Simultaneously, grouped sequential control avoids compressor load peaks caused by multiple evaporators defrosting simultaneously. Combined with cloud-based global scheduling and time-of-use pricing for off-peak operation, it effectively reduces system peak power consumption and overall electricity costs.

[0060] This invention's precise frost risk prediction ensures timely intervention before frost affects heat exchange efficiency, preventing temperature fluctuations and reduced cooling capacity due to excessive frost thickness. Multi-parameter safety monitoring during the defrosting process (discharge pressure, superheat, surface temperature) effectively prevents liquid slugging and high-pressure failures, ensuring safe compressor operation, maintaining constant storage / cabinet temperatures, and guaranteeing food storage quality.

[0061] This invention features a local controller with edge computing capabilities that can operate independently during communication interruptions, ensuring system reliability. A cloud-based centralized control platform aggregates data from multiple sites, providing real-time monitoring, energy consumption reports, and multi-site management functions through a visualization module, significantly reducing manual inspection costs. The predictive maintenance module analyzes equipment health status based on long-term data, transforming reactive maintenance into proactive, early warning-based maintenance.

[0062] This invention integrates the prediction model and uses the actual time and energy consumption data of each defrost as feedback, uploads it to the cloud for iterative optimization, realizes the "self-learning" capability of the control strategy to become more and more accurate with use, and enables the system to adapt to complex operating conditions such as seasonal changes and changes in usage habits, and maintain the optimal operating state in the long term. Specific Implementation Example 1: Centralized control system applied to chain supermarkets A supermarket chain has 10 stores, each containing 20 sets of refrigerated display cases (including refrigerators and freezers). The system hardware is deployed according to the above configuration, and the cloud platform adopts a private cloud architecture.

[0064] Data Acquisition: Each display case is equipped with a temperature sensor, differential pressure transmitter, and door magnetic switch. The PLC collects data every 10 seconds, calculates features such as ΔT, ΔP, and door_1h, and uploads the data to the cloud every minute.

[0065] Rule-based scoring model: The PLC operates according to scoring rules, with thresholds set as follows: ΔT < 2.5℃ (weight 3), ΔP > 120Pa (weight 3.5), RH > 80% (weight 2), t_since > 180min (weight 2), and door_1h > 20min (weight 1.5). The high-risk threshold is set to 7 points. Operating data for a freezer: ΔT = 2.0℃, ΔP = 150Pa, RH = 75%, t_since = 200min, door_1h = 15min. The calculated score is: 3 + 3.5 + 0 + 2 + 0 = 8.5 > 7, and the PLC immediately triggers defrosting.

[0066] Grouping sequence control: Since three groups of evaporators need to defrost at the same time, the PLC divides them into three groups, A, B and C, and executes them in the order of A→B→C. The defrosting interval between each group is 10 minutes to ensure stable compressor load.

[0067] Cloud-based optimization: After collecting data from all stores in the cloud, a reinforcement learning model is used to generate a defrosting plan for the next day. High-risk defrosting is scheduled during off-peak hours (23:00-7:00) at night, and the adjusted trigger thresholds are sent to the PLCs of each store.

[0068] Results: After implementation, compared with traditional timed defrosting, the system's total energy consumption was reduced by 22%, the temperature fluctuation of the storage room / cabinet was controlled within ±1.2℃, the number of defrosting cycles was reduced by 30%, and the maintenance early warning accuracy rate reached 85%.

[0069] Example 2: Independent Control System for Small Cold Storage A single cold storage unit (capacity 500m³) 3This simplified version of the invention is equipped with an integrated touchscreen controller (integrating edge computing and HMI), eliminating the need for cloud connectivity. The controller incorporates a rule-based scoring model and a lightweight random forest model (pre-trained based on historical data). It directly controls the hot gas defrosting pipeline, supporting local parameter settings and status monitoring. When the random forest model outputs a defrosting probability > 0.8, the defrosting sequence is automatically executed. Cold storage managers can view defrosting records and alarm information via a mobile app (direct Bluetooth connection). This solution reduces costs and enables intelligent, on-demand defrosting.

[0070] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0071] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A centralized control method of hot gas defrosting, characterized by, Includes the following steps: Step 1, Data Acquisition Step: Real-time acquisition of multi-dimensional operating parameters of the refrigeration system, including at least the evaporator inlet and outlet temperature difference, evaporator differential pressure, ambient relative humidity, opening status of the refrigeration unit door or cabinet door, evaporator surface temperature, compressor exhaust parameters, and system energy consumption. Step 2, Frost Prediction and Triggering Step: Based on the collected multi-dimensional operating parameters, the fusion prediction model is used to dynamically assess the frost risk of each evaporator. When the frost risk reaches the preset triggering condition, a defrosting command is generated. The fusion prediction model includes a rule-based fast scoring model and / or a machine learning-based accurate prediction model. Step 3, Defrosting Execution Steps: In response to the defrosting command, control the target evaporator to execute a hot gas defrosting sequence, which includes: shutting off the liquid supply, stopping or reducing the fan speed, opening the hot gas bypass valve, monitoring the defrosting process parameters, and closing the hot gas valve and restoring cooling when the termination conditions are met. Step 4: Optimize the feedback process: Record the actual time and energy consumption of the defrosting process, and upload it as feedback data to the cloud-based centralized control platform for updating and optimizing the fusion prediction model.

2. The centralized control method of hot gas defrosting according to claim 1, characterized in that: The rule-based fast scoring model is specifically as follows: Based on multiple preset risk factors and their weights, the current real-time data is weighted and scored to obtain a risk score. The risk factors include whether the temperature difference between the inlet and outlet of the evaporator exceeds the first temperature difference threshold, whether the differential pressure across the evaporator exceeds the first differential pressure threshold, whether the relative humidity of the environment exceeds the first humidity threshold, whether the time since the end of the last defrost exceeds the first time threshold, and whether the cumulative time of door opening within the preset period exceeds the first time threshold. The risk score is compared with a preset risk level threshold. If it exceeds the high risk threshold, it is determined to be high risk and a defrosting command is immediately triggered. If the risk exceeds the medium-risk threshold but does not reach the high-risk threshold, it is judged as medium-risk, and the frequency of data collection and monitoring is increased.

3. The centralized control method of hot gas defrosting according to claim 2, characterized in that: The machine learning-based accurate prediction model is a long short-term memory network model or a random forest model; the LSTM model takes historical time series data as input and outputs a predicted value for the future frost growth rate or frost probability; the RF model takes the multidimensional features and statistical lag features of the current moment as input and outputs the frost melting probability; the trigger condition is that the predicted frost growth rate or frost probability exceeds a preset machine learning trigger threshold.

4. The centralized control method of hot gas defrosting according to claim 1, characterized in that: The defrosting execution step also includes a grouping sequence control sub-step: when multiple evaporators need to be defrosted at the same time, the evaporators to be defrosted are divided into multiple defrosting groups. At the same time, only one group or a preset number of defrosting groups are allowed to execute the hot gas defrosting sequence at the same time, while the remaining defrosting groups are in a waiting state.

5. The centralized control method of hot gas defrosting according to claim 1, characterized in that: The defrosting process also includes a safety monitoring sub-step: During hot gas defrosting, the evaporator surface temperature, system exhaust pressure, and suction superheat are monitored in real time. When the evaporator surface temperature rises to the preset defrosting completion temperature threshold, the defrosting is considered to be over. When the exhaust pressure exceeds the safety limit or the suction superheat is lower than the liquid slugging risk threshold, the current defrosting process is immediately interrupted and an alarm signal is issued.

6. The centralized control method of hot gas defrosting according to claim 1, characterized in that: The optimization feedback step further includes a cloud-based centralized optimization sub-step: the cloud-based centralized control platform aggregates data uploaded from multiple refrigeration stations, combines time-of-use electricity price information and frost risk prediction for each station, uses an optimization algorithm to generate a global defrosting scheduling plan, and sends the adjusted control parameters to the local controller.

7. A centralized control system for hot gas defrosting, for implementing the method according to any one of claims 1 to 6, characterized in that, include: Sensor arrays are deployed at various monitoring points in the refrigeration system to collect multi-dimensional operating parameters in real time. An edge data acquisition module, connected to the sensor group, is used for preprocessing and feature extraction of the acquired data; A local controller, connected to the edge data acquisition module, is used to execute local control logic, including receiving defrosting commands and controlling the actuator to complete the hot gas defrosting sequence; The communication module is used to enable data exchange between the local controller and the cloud-based centralized control platform; A cloud-based centralized control platform, connected to the communication module, is used to store and analyze all uploaded data, run the fusion prediction model, generate and issue defrosting commands, and perform global optimization scheduling. The actuator, connected to the local controller, includes at least a hot gas solenoid valve, a liquid supply solenoid valve, an electronic expansion valve, and a fan, and is used to perform corresponding switching or regulating actions in response to control signals.

8. The centralized control system for heat gas defrosting according to claim 7, wherein The sensor group includes at least: Temperature sensors used to measure the inlet and outlet temperatures of an evaporator; Differential pressure transmitter used to measure the differential pressure across an evaporator; A humidity sensor used to measure the relative humidity of the environment; Door magnetic switches or infrared sensors used to monitor the opening status of warehouse or cabinet doors; Temperature sensor used to measure the surface temperature of an evaporator; Pressure transmitters and temperature sensors used to measure compressor discharge pressure and temperature; An energy meter used to measure system energy consumption.

9. The centralized control system for heat gas defrosting according to claim 7, wherein The local controller is a programmable logic controller or an embedded controller with edge computing capabilities. It can independently run the rule-based fast scoring model and maintain local automatic control when communication with the cloud is interrupted.

10. The centralized control system for heat gas defrosting according to claim 7, wherein The cloud-based centralized control platform also integrates a data visualization module and a predictive maintenance module. The data visualization module provides real-time / historical data queries, energy consumption reports, and a multi-site group management interface. The predictive maintenance module assesses the health status of equipment and issues maintenance warnings based on long-term data analysis.

Images