Intelligent refrigerant compressor system based on permanent magnet synchronous motor and Internet of Things and control method

Abstract

The invention relates to the technical field of refrigeration equipment, in particular to an intelligent refrigerant compressor system based on a permanent magnet synchronous motor and the Internet of Things and a control method, and the system comprises a permanent magnet synchronous motor module, an intelligent control module, an Internet of Things module and a multi-machine group control module, wherein the permanent magnet synchronous motor module adopts a Halbach array, double cooling systems and magnetic steel health monitoring structure, and is combined with an NdFeB magnetic steel nano anticorrosive coating and magnetic flux density real-time detection, so that the efficiency is greater than or equal to 96%, and demagnetization is prevented; the intelligent control module is based on'load-environment collaborative prediction dynamic weight model prediction control (MPC) ', and integrates environment temperature, refrigerant pressure prediction and vibration suppression algorithms to realize full-working-condition energy efficiency optimization; the Internet of Things module constructs an edge-cloud collaborative architecture, integrates refrigerant leakage traceability of vibration voiceprint analysis and LSTM fault early warning, and supports 4G + LoRaWAN dual-mode communication; and the multi-unit group control module realizes load distribution optimization of units within 256 through an energy efficiency balance scheduling algorithm.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of refrigeration equipment, in particular to an intelligent refrigerant compressor system based on a permanent magnet synchronous motor and the Internet of Things and a control method thereof. BACKGROUND

[0002] As the core power component of a refrigeration system, the performance of a refrigerant compressor directly determines the energy efficiency, reliability and operating cost of the refrigeration equipment. However, the existing refrigerant compressor system still has many defects to be solved in the technical application and actual operation, which can be analyzed from the following dimensions: Firstly, in the power driving link, the traditional refrigerant compressor generally adopts an asynchronous induction motor, and its rated efficiency is usually less than 90%, and the efficiency sharply decreases under some load conditions (such as commercial scenarios with fluctuating refrigeration demand) - when the load rate decreases to 50%, the efficiency of the induction motor is usually less than 65%, resulting in a large amount of energy waste. For example, a traditional induction motor compressor is used in a central air conditioning system of a shopping mall, and during the night low load period (load rate 40%-50%), the daily energy consumption of a single compressor is only reduced by 15% compared with the rated load, which is far lower than the load rate reduction, and the annual additional energy consumption can reach 12,000 kWh. At the same time, the traditional motor adopts a single cooling method (such as natural air cooling or simple oil cooling), which cannot adapt to the temperature rise demand under high load - when the motor runs at a temperature above 120℃ for a long time, the insulation performance of the winding decreases by 30%, and the service life of the motor is shortened to 60% of the designed service life; in addition, the motor rotor magnetic steel (mostly NdFeB material) lacks effective health monitoring and protection mechanism, and is easy to demagnetize in high temperature (>150℃) or corrosive environment, once the magnetic flux density of the magnetic steel decreases by ≥8%, the efficiency of the motor will decrease by more than 25%, and it cannot be repaired by conventional maintenance, and the entire rotor needs to be replaced, increasing the maintenance cost by tens of thousands of yuan.

[0003] Secondly, at the control strategy level, existing compressors mostly use fixed-frequency control or simple variable-frequency control, lacking the ability to cooperatively predict loads and environments. On the one hand, fixed-frequency compressors adjust the refrigeration capacity through "start-stop cycles", and the impact current during the start-stop process can reach 5-8 times the rated current, which not only causes an impact on the power grid, but also leads to increased mechanical wear of the compressor - statistics show that the start-stop loss of fixed-frequency compressors accounts for 15%-20% of the total energy consumption, and for every 1000 times the start-stop frequency increases, the service life of the mechanical parts of the compressor is reduced by 10%. On the other hand, conventional variable-frequency control only adjusts the frequency according to the real-time load rate, without considering the influence of environmental factors (such as ambient temperature, humidity, and refrigerant leakage) on the load. For example, during the summer high-temperature period (ambient temperature > 35°C), the refrigerant pressure rises sharply, causing a load fluctuation amplitude of up to 30%. The response lag time of conventional variable-frequency control is > 500 ms, which easily leads to "overcooling" or "undercooling" phenomena, and the refrigeration efficiency is reduced by 12%-15%. In addition, existing vibration suppression techniques mostly use passive damping (such as rubber damping pads), which have poor suppression effect on the 100-500 Hz high-frequency vibrations generated during motor operation. High-frequency vibrations not only cause noise to rise (up to more than 80 dB), but also exacerbate bearing wear. The fault detection rate of bearings is only about 65%, and irreversible damage is often caused when the fault is detected, resulting in an average of 48 hours / stoppage of unexpected downtime.

[0004] Furthermore, at the monitoring and operation stage, traditional compressors rely on manual inspection to achieve fault early warning and data recording, which has problems such as response lag and fragmented data. On the one hand, the manual inspection cycle is usually 1 week to 1 month, which cannot capture sudden faults (such as refrigerant leakage and winding short circuit) in real time. For example, during the initial stage of refrigerant leakage (leakage amount < 0.5 kg / h), there are no obvious symptoms, and manual inspection cannot detect it. By the time the refrigeration capacity decreases, the leakage has been going on for 10-15 days, not only wasting refrigerant (the annual loss can reach 20% of the total refrigerant capacity), but also damaging the ozone layer. On the other hand, the precision of manually recorded operating data (such as current and temperature) is low (error ± 5%), and unified analysis of multi-unit data cannot be achieved, resulting in that multi-unit control can only achieve "simple start-stop control", and cannot allocate loads according to the energy efficiency differences of each unit. In a certain cold-chain logistics park, the 10 traditional compressor units of the group control system were running in an uneven load distribution, with some units running at over 95% overload and some units in low-load standby, resulting in a 20% decrease in overall energy efficiency compared to the ideal state, and the failure frequency of overloaded units was 3 times that of normal units.

[0005] Finally, at the level of multi-machine cooperation, the existing group control system lacks an energy efficiency balanced scheduling mechanism. When multiple compressors are running in parallel, the traditional group control only allocates the load according to the "first start first stop" or "fixed sequence", ignoring the real-time energy efficiency difference of each unit (such as the COP value of new machines can reach 4.5, while the COP value of old machines running for 3 years drops to 3.8), resulting in long-term full-load operation of high-energy-consumption units, while high-energy-efficiency units are underloaded, and the overall system energy efficiency is wasted. At the same time, the networking scale of the existing group control system is limited (multiple < 100 units), and the communication stability is poor - when using single 4G communication, the packet loss rate in weak signal areas (such as underground machine rooms) can reach 5%-10%, causing data transmission interruption and group control failure.

[0006] In summary, the existing refrigerant compressor system has significant defects in power efficiency, control accuracy, monitoring and operation, multi-machine cooperation, etc., and urgently needs a technical solution integrating efficient permanent magnet drive, intelligent predictive control, Internet of Things real-time monitoring and multi-machine energy efficiency balancing to solve the above problems. SUMMARY

[0007] The purpose of the present application is to overcome the defects of the prior art and provide an intelligent refrigerant compressor system and control method based on permanent magnet synchronous motor and Internet of Things, which realizes high energy efficiency, high reliability and intelligent management of the compressor system through multi-module collaborative innovation.

[0008] The technical scheme adopted by the present application to solve its technical problems is: an intelligent refrigerant compressor system and control method based on a permanent magnet synchronous motor and an Internet of Things, comprising a permanent magnet synchronous motor module, an intelligent control module, an Internet of Things module that are sequentially electrically connected and interact in data, and a multi-machine group control module in communication connection with the Internet of Things module; the permanent magnet synchronous motor module comprises a rotor, a stator, a double cooling system and a magnetic steel health monitoring unit, the rotor adopts a Halbach array and a radial segmented hybrid magnetic circuit, the surface of the magnetic steel is covered with an Al2O3 nano anti-corrosion coating, the double cooling system comprises an oil cooling channel (provided in the interior of the stator core in a spiral shape) and a wind cooling auxiliary device (provided in the motor end cover and containing a temperature self-adaptive fan), the magnetic steel health monitoring unit contains a PT100 temperature sensor and a magnetic flux density detection coil (embedded in the gap between the rotor magnetic steels); the intelligent control module comprises a load-environment prediction unit, a dynamic weight MPC algorithm unit and a vibration suppression unit, the load-environment prediction unit predicts the load change based on the data of an environmental temperature sensor (accuracy ±0.5℃) and a refrigerant high pressure sensor (range 0-4MPa), the dynamic weight MPC algorithm unit adjusts the weight coefficients (ω1, ω2, ω3) of torque, flux linkage and energy consumption through fuzzy logic, and the vibration suppression unit adopts FFT analysis and active damping algorithm; the Internet of Things module comprises an edge computing layer (adopting Xilinx Zynq UltraScale+MPSoC and supporting real-time FFT and voiceprint analysis), a network layer (4G+LoRaWAN dual-mode communication, LoRaWAN spreading factor SF=12) and a cloud platform layer (containing a digital twin model and a refrigerant leakage tracing module, leakage positioning accuracy ±0.5m); the multi-machine group control module is based on an energy efficiency balanced scheduling algorithm, allocates loads according to the real-time COP values of each unit to avoid single unit overload (load rate ≤95%).

[0009] Specifically, the stator of the permanent magnet synchronous motor module adopts 12-slot 10-pole slot-pole matching, the ratio of q-axis inductance to d-axis inductance is ≥2.2, the rated power is 50-315kW, the speed regulation range is 500-3500rpm, and the peak torque is 200-2000N・m.

[0010] Specifically, the load-environment prediction unit of the intelligent control module adopts a BP neural network prediction model, the input parameters include the average value of the environmental temperature in the past 10min, the refrigerant high pressure fluctuation value and the compressor operating frequency, the output is the load prediction value in the future 5min, and the prediction error is ≤3%.

[0011] Specifically, the edge computing layer of the Internet of Things module further integrates a multi-scale entropy analysis unit for processing bearing vibration data collected by a vibration sensor (range 0-50g, frequency response 0-10kHz), and realizes a bearing fault recognition rate of >98% in combination with a CNN diagnosis model.

[0012] Specifically, the energy efficiency balancing scheduling algorithm of the multi-machine group control module includes the following steps: ①collecting real-time COP values, load rates, and winding temperatures of each machine group; ②calculating the energy efficiency margin (actual COP value / rated COP value) of each machine group; and ③allocating the newly added load to the machine group with the highest energy efficiency margin until its load rate is ≤95%.

[0013] Specifically, the oil cooling channel of the double cooling system adopts a topology structure optimized by ANSYS Fluent, the oil flow is dynamically adjusted according to the stator temperature (the oil flow increases by 30% when the temperature ≥120℃), and the fan speed of the air cooling auxiliary device is positively correlated with the motor end cover temperature (the speed increases by 15% when the temperature increases by 10℃).

[0014] Specifically, the digital twin model of the cloud platform layer is built based on an InfluxDB time series database, the data update period is ≤3s, the life prediction adopts a Wiener process degradation modeling, and the prediction accuracy is >88%.

[0015] Specifically, the magnetic flux density detection coil output voltage signal of the magnetic steel health monitoring unit is transmitted to the intelligent control module after AD conversion (16-bit precision), and when a decrease of ≥5% in the magnetic flux density is detected, the magnetic steel protection mode is triggered (reducing the motor output torque by 10%-20%).

[0016] The intelligent refrigerant compressor control method based on a permanent magnet synchronous motor and an Internet of Things includes the following steps: ①Initialization: the permanent magnet synchronous motor module is powered on, the Internet of Things module establishes edge-cloud communication, and the multi-machine group control module reads the initial state of each machine group; ②Load prediction: the intelligent control module collects environmental temperature and refrigerant pressure data, and outputs a 5min load prediction value through a BP neural network; ③Mode switching: the dynamic weight MPC algorithm unit adjusts the weight coefficient according to the load prediction value (ω3>ω1 when the load <70%, ω1=ω3 when the load is 70%-95%, and ω1>ω3 when the load >95%), and switches to the economic / standard / overload mode; ④Real-time regulation and control: the vibration suppression unit analyzes the motor vibration frequency through FFT, and outputs an active damping control signal; the double cooling system adjusts the oil flow and fan speed according to the stator temperature and end cover temperature; ⑤Fault early warning and leakage tracing: the edge computing layer analyzes the vibration and magnetic flux data, the cloud platform layer early warns the fault through an LSTM model for 72h, and locates the refrigerant leakage point combined with the voiceprint data; and ⑥Multi-machine scheduling: the multi-machine group control module allocates the load according to the energy efficiency margin of each machine group, and realizes the overall energy efficiency optimization.

[0017] Specifically, the FFT analysis frequency range of the vibration suppression unit in step ④ is 0-2kHz, and the response time of the active damping control is ≤10ms; and the refrigerant leakage tracing in step ⑤ adopts voiceprint feature matching (the feature extraction includes sound pressure level and frequency spectrum peak), and the positioning time is ≤10s.

[0018] The beneficial effects of the present application are: Improve power drive efficiency and reliability: The permanent magnet synchronous motor module of the system adopts Halbach array and radial segmented mixed magnetic circuit design, optimizes the air gap flux density distribution, greatly improves the motor operating efficiency; matched with Al2O3 nano anti-corrosion coating, effectively resist the corrosion of refrigerant, protect the performance of magnetic steel; double cooling system can dynamically adjust the cooling intensity according to the motor temperature, avoid the damage of high temperature to the motor parts; magnetic steel health monitoring unit can track the state of magnetic steel in real time, trigger the protection mechanism in advance, prevent the demagnetization of magnetic steel, significantly prolong the service life of motor, at the same time ensure the stable operation of motor in wide load range, reduce the downtime risk caused by power component failure.

[0019] Optimize control accuracy and energy consumption performance: Intelligent control module can predict the load change trend in advance through load-environment collaborative prediction, reserve sufficient time for control strategy adjustment, avoid the problem of "overcooling", "undercooling" or response lag in traditional control; dynamic weight MPC algorithm can flexibly adjust the control focus according to the load condition, balance the torque output and energy consumption optimization in different operation modes, reduce energy waste; active vibration suppression technology can accurately identify and suppress high-frequency vibration in motor operation, not only reduce the operation noise, but also reduce the wear of bearing and other mechanical parts, further improve the system operation stability and reduce the long-term maintenance cost.

[0020] Realize intelligent monitoring and efficient operation: The edge-cloud collaborative architecture built by the Internet of Things module can process motor vibration, magnetic flux, voiceprint and other data in real time, the edge computing layer can quickly complete local fault preliminary judgment and emergency control, the cloud platform layer realizes system operation state visualization through digital twin model, and finds potential faults in advance with the help of fault warning model, so as to give maintenance personnel sufficient repair time and reduce unexpected downtime; refrigerant leakage tracing function can quickly locate the leakage point, avoid refrigerant waste and environmental damage, and prevent refrigerant leakage from reducing refrigeration efficiency, ensuring stable refrigeration effect, especially suitable for scenes sensitive to refrigerant leakage.

[0021] Improve multi-machine group collaborative energy efficiency and compatibility: The energy efficiency balancing algorithm of multi-machine group control module can allocate load according to the real-time energy efficiency difference of each machine group, preferentially utilize the potential of high energy efficiency machine group, avoid the situation that some machine groups are overloaded and some machine groups are in low load standby, and improve the overall operation energy efficiency of multi-machine group; The system supports large-scale machine group networking, and is compatible with compressors of different brands and different powers, has strong adaptability, can meet the demand of multi-machine group collaborative operation in large commercial air conditioning, cold chain logistics and other scenes, and reduce the machine group loss caused by uneven load distribution, prolong the service life of the whole group control system.

[0022] Enhance scene adaptability and operation stability: For different refrigeration scenes (such as wide temperature fluctuation of commercial central air conditioner, low temperature continuous operation of cold chain logistics), the system optimizes the hardware configuration and algorithm design, such as low temperature adaptability of magnetic steel, compensation of load prediction influence of warehouse door switch, etc., to ensure stable operation under different environmental conditions; At the same time, through multiple protection mechanisms (such as high temperature protection, overload protection, low temperature insulation protection, etc.), the system fault probability under extreme working conditions is further reduced, the overall reliability is improved, and the economic loss caused by scene adaptation problem is reduced. DETAILED DESCRIPTION

[0023] In order to make the technical means, creative characteristics, purposes and effects realized by the present application easy to understand, the present application will be further described below in combination with specific embodiments.

[0024] The intelligent refrigerant compressor system and control method based on permanent magnet synchronous motor and Internet of Things described in the present application have the following technical solutions: Permanent magnet synchronous motor module: high-efficiency drive design with double cooling + magnetic steel health monitoring This module is the power core of the system. In view of the problems of low efficiency, easy demagnetization of magnetic steel and insufficient cooling of traditional motor, an integrated structure of "Halbach array + double cooling + magnetic steel health monitoring" is adopted, and the specific design is as follows: Rotor magnetic circuit optimization: the rotor adopts Halbach array and radial segmented mixed magnetic circuit. Halbach array can make the air gap magnetic flux density distribution closer to sine wave, reduce 5th and 7th harmonic content (harmonic distortion rate ≤3%), and improve motor efficiency. The radial segmentation divides the magnetic steel into 3 segments (each segment is 15mm thick), and a 0.5mm thick non-magnetic gasket is arranged between the segments to avoid thermal stress concentration of the magnetic steel due to temperature gradient, while reducing eddy current loss (eddy current loss is reduced by 40% compared with the whole magnetic steel). The magnetic steel adopts N38SH type NdFeB material, and the surface is covered with 5μm thick Al2O3 nano anti-corrosion coating prepared by magnetron sputtering process, the porosity of the coating is <0.1%, which can resist the corrosion of refrigerant (such as R32, R410A), and solve the corrosion problem of traditional magnetic steel in refrigerant environment (corrosion rate is reduced from 0.2mm / year to 0.01mm / year).

[0025] Dual cooling system: including spiral oil cooling channel and temperature adaptive air cooling device. The spiral oil cooling channel is arranged inside the stator core, the channel diameter is 8 mm, the pitch is 20 mm, and ANSYS Fluent is used for topology optimization. Simulation shows that the oil cooling channel can reduce the stator temperature by 25°C under rated load, and the oil flow can be dynamically adjusted according to the stator temperature (the PWM signal output by the intelligent control module controls the oil pump, and when the stator temperature is greater than or equal to 120°C, the oil flow increases from 20 L / min to 26 L / min); the air cooling auxiliary device is installed on the non-shaft extension end cover of the motor, which contains a 120 mm diameter axial fan, and the fan speed is controlled by the end cover temperature sensor (PT100, accuracy ±0.1°C) — when the end cover temperature is less than or equal to 80°C, the fan stops, when the temperature is between 80°C and 100°C, the speed is 1500 rpm, and when the temperature is greater than 100°C, the speed is 2500 rpm, realizing "on-demand cooling" and avoiding energy waste of traditional single cooling (cooling energy consumption is reduced by 30%).

[0026] Magnetic steel health monitoring unit: composed of PT100 temperature sensor and magnetic flux density detection coil. The PT100 sensor is embedded in the groove (depth 2 mm) on the surface of the rotor magnetic steel, which can collect the temperature of the magnetic steel in real time, and trigger a high-temperature warning (send alarm information through the Internet of Things module) when the temperature is greater than 140°C; the magnetic flux density detection coil (enameled wire diameter 0.1 mm, number of turns 500 turns) is embedded in the gap between the rotor magnetic steel, and the coil output voltage signal is proportional to the magnetic flux density (sensitivity 0.5V / T), the signal is converted by a 16-bit AD converter (sampling rate 1 kHz) and transmitted to the intelligent control module, when the detected magnetic flux density decreases by more than 5% (corresponding to the initial stage of magnetic steel demagnetization), the intelligent control module automatically reduces the motor output torque by 10%-20%, and starts the magnetic steel protection mode (such as increasing the cooling oil flow), to avoid further demagnetization of the magnetic steel — experimental data shows that this protection mechanism can reduce the demagnetization rate of the magnetic steel by 80%, and the service life of the motor can be extended to more than 15 years.

[0027] In addition, the stator adopts 12-slot 10-pole slot-pole matching, which can eliminate 6k±1 harmonic (k is a positive integer) and reduce torque ripple (torque ripple rate ≤2%); the motor rated power covers 50-200kW, the speed range is 500-3500rpm, the peak torque is 200-500N・m, the q-axis inductance and d-axis inductance ratio is greater than or equal to 2.2, which ensures the high efficiency of the motor in a wide load range (full working condition efficiency >92%, rated load efficiency ≥96%).

[0028] Intelligent control module: dynamic weight MPC design of load-environment collaborative prediction This module is the control core of the system, in order to solve the problems of response lag and insufficient energy efficiency optimization of traditional control, a three-layer control architecture of "load-environment collaborative prediction + dynamic weight MPC + active vibration suppression" is adopted, the specific design is as follows: Load-environment collaborative prediction unit: a prediction model is constructed based on BP neural network, the input parameters include "average value of environmental temperature in the past 10 minutes (collected by DS18B20 temperature sensor installed outside the unit, accuracy ±0.5℃), refrigerant high pressure fluctuation value (collected by MPX5700 pressure sensor installed at the exhaust port of the compressor, range 0-4MPa, accuracy ±1%FS), compressor operating frequency", and the output is "load prediction value in the next 5 minutes". The model training uses 1000 sets of historical operation data (covering environmental temperature -10℃-45℃, load rate 30%-100%), and the prediction error is ≤3% after training - for example, when the environmental temperature rises from 30℃ to 35℃ and the refrigerant high pressure rises from 2.5MPa to 3.2MPa, the model can predict that the load rate rises from 60% to 85% in advance for 5 minutes, reserving time for control strategy adjustment.

[0029] Dynamic weight MPC algorithm unit: the target function of MPC (model predictive control) is , wherein is the reference torque, is the estimated torque, is the flux amplitude, is the total motor loss (including copper loss, iron loss, mechanical loss). The weight coefficients ω1 (torque tracking weight), ω2 (flux control weight), and ω3 (energy optimization weight) are dynamically adjusted through fuzzy logic: when the load rate is <70% (economic mode), ω3=0.6, ω1=0.3, and ω2=0.1, the energy consumption is preferentially reduced; when the load rate is 70%-95% (standard mode), ω1=0.4, ω3=0.4, and ω2=0.2, torque tracking and energy consumption are balanced; when the load rate is >95% (overload mode), ω1=0.6, ω2=0.3, and ω3=0.1, the torque output is preferentially guaranteed. The algorithm solves the optimal voltage vector V dq through quadratic programming (QP), and then outputs to the motor inverter through space vector pulse width modulation (SVPWM), with a response time of ≤50ms - experiments show that compared with traditional PI control, the motor energy consumption is reduced by 15%, and the torque tracking error is reduced from 5% to 1.5%.

[0030] Active vibration suppression unit: combined scheme of "FFT analysis + active damping". Vibration sensor (ADXL357, range 0-50g, frequency response 0-10kHz) is installed on the motor bearing end cover to collect vibration acceleration signals; intelligent control module performs FFT analysis (frequency range 0-2kHz) on the signals to identify the main vibration frequency (such as 100Hz electromagnetic vibration, 300Hz mechanical vibration); active damping algorithm outputs compensation current (through inverter to inject motor stator winding) according to vibration frequency, generates electromagnetic force opposite to vibration direction, suppresses vibration - tests show that this unit can reduce motor vibration acceleration from 0.8g to 0.2g, noise from 78dB to 70dB, and bearing wear rate by 30%.

[0031] Internet of Things module: edge-cloud collaborative monitoring and traceability design This module is the intelligent core of the system. To solve the problems of traditional monitoring lag and difficulty in locating leaks, a three-level architecture of "edge computing layer-network layer-cloud platform layer" is constructed, with the following specific design: Edge computing layer: Xilinx Zynq UltraScale+MPSoC chip (including ARM Cortex-A53 processor and FPGA) is used to realize real-time data processing and local decision-making. Core functions include: ① vibration data processing: multi-scale entropy analysis is performed on bearing vibration signals (extracting 6 feature parameters: entropy, mean, variance, peak, kurtosis, skewness), combined with CNN diagnostic model (input feature vector, output fault type: normal, outer ring wear, inner ring wear, ball damage), fault recognition rate > 98%, recognition time ≤ 1s; ② voiceprint data collection: through the microphone (frequency response 20Hz-20kHz) installed on the compressor shell, the running voiceprint is collected, when the refrigerant leakage occurs (leakage amount ≥ 0.1kg / h), the voiceprint will appear characteristic frequency of 2000-3000Hz, the edge computing layer captures this frequency in real time and preliminarily judges whether the leakage exists; ③ local emergency control: when the winding temperature > 150℃ or the bearing vibration exceeds the limit (> 1g), without cloud platform instruction, directly output stop signal, response time ≤ 100ms, avoid accident expansion.

[0032] Network layer: 4G+LoRaWAN dual-mode communication is adopted to ensure communication reliability in different scenarios. 4G communication is used for large data transmission (such as digital twin model update, historical data upload), with transmission rate ≥1 Mbps and delay ≤100 ms; LoRaWAN communication is used for small data transmission and long-distance transmission (such as real-time state parameters, alarm information), with spreading factor SF=12, communication distance up to 5 km (open environment), and packet loss rate <0.1% (conflicts are reduced through time slot ALOHA protocol). Dual-mode communication adopts a "primary and backup switching" mechanism: when the 4G signal strength is < -90 dBm, it automatically switches to LoRaWAN to ensure uninterrupted communication - tests show that in the underground room (4G signal weak) scenario, the packet loss rate of dual-mode communication is reduced from 8% of single 4G to 0.05%.

[0033] Cloud platform layer: based on Kubernetes cluster and InfluxDB time series database, the core functions include: ① Digital twin model: 1:1 restoration of compressor system structure and operating state, real-time synchronization of 16 parameters such as motor speed, winding temperature, refrigerant pressure, data update period ≤3s, supporting remote visual monitoring (such as viewing temperature distribution of each component through Web); ② Fault warning: using LSTM neural network (input past 24h vibration, temperature, current data, output future 72h fault probability), warning accuracy >96%, such as predicting bearing outer ring wear failure 48h in advance; ③ Refrigerant leakage tracing: combined with soundprint data from edge computing layer and location information of multiple units (collected through GPS module, accuracy ±1m), soundprint propagation model is constructed, locating the leakage point with accuracy ±0.5m and tracing time ≤10s - experiments show that this function can shorten the refrigerant leakage detection time from 10 days to 10 seconds, reducing refrigerant loss by 90% per year.

[0034] Multi-unit control module: collaborative design of energy efficiency balanced scheduling This module adopts energy efficiency balanced scheduling algorithm to realize optimal collaboration of multiple units to solve the problem of uneven load distribution in traditional group control, with the following specific design: Energy efficiency margin calculation: the multi-unit control module collects real-time data (COP value, load rate, winding temperature, running time) of each unit through the Internet of Things module, calculates the "energy efficiency margin" of each unit, and the formula is: , where is the ratio of real-time refrigeration capacity to input power, is the rated COP of the unit when it leaves the factory (new machine takes 4.5, runs for 1 year takes 4.3, runs for 2 years takes 4.1, runs for 3 years takes 3.8). The higher the energy efficiency margin, the greater the energy efficiency potential of the unit, and the more suitable it is to bear new loads.

[0035] Load distribution strategy: When the system needs to increase the refrigeration capacity, the multi-machine group control module allocates the load in the order of "energy efficiency margin from high to low", and the load increment is 5% each time until the new load is fully allocated or the unit load rate reaches 95% (overload threshold). When the system needs to reduce the refrigeration capacity, the load is reduced in the order of "energy efficiency margin from low to high" to avoid premature shutdown of high energy efficiency units. For example, in a certain scenario, there are 3 units: unit A (COP actual = 4.2, load rate 60%, energy efficiency margin = 0.42 x 0.4 = 0.168), unit B (COP actual = 3.9, load rate 80%, energy efficiency margin = 0.9 x 0.2 = 0.18), unit C (COP actual = 3.7, load rate 70%, energy efficiency margin = 0.97 x 0.3 = 0.291), when the new load is 10%, it is preferentially allocated to unit C (5%, load rate 75%), then to unit B (5%, load rate 85%), and unit A is not allocated, ensuring the optimal overall energy efficiency.

[0036] Networking scale and compatibility: Supports networking of up to 256 units, uses Modbus-RTU protocol to realize data interaction between units, and is compatible with compressors of different brands and powers (as long as they have a 485 communication interface). Tests show that when 256 units are networked, the transmission delay of group control instructions is ≤500ms, the load distribution completion time is ≤10s, and the overall system energy efficiency is improved by 20% compared with traditional group control.

[0037] Example 1: Scroll-type intelligent refrigerant compressor system for commercial central air conditioning This embodiment is aimed at the commercial central air conditioning scene (such as large shopping malls, office buildings), which is characterized by large fluctuations in environmental temperature (30-40℃ in summer, 5-15℃ in winter) and frequent changes in load (80-100% during the day, 30-50% at night), with high requirements for compressor energy efficiency, response speed and reliability. The specific implementation details are as follows: System hardware configuration Permanent magnet synchronous motor module: uses a 110kW scroll-type compressor dedicated permanent magnet synchronous motor with a rated speed of 1500rpm, a peak torque of 380N・m, a speed regulation range of 600-3000rpm, and a q-axis inductance to d-axis inductance ratio of 2.3; the rotor magnetic steel is N38SH type NdFeB with a surface Al2O3 nano coating thickness of 5μm; the oil cooling channel of the double cooling system has a diameter of 8mm and a pitch of 20mm, and the oil pump model is CB-B100 (maximum flow 100L / min), and the air cooling fan is a 120mm axial flow fan (model AFB1212H); the PT100 sensor of the magnetic steel health monitoring unit is model WZP-PT100 (accuracy ±0.1℃), and the magnetic flux density detection coil output voltage range is 0-5V (corresponding to magnetic flux density 0-10T).

[0038] Intelligent control module: STM32H743 microcontroller (main frequency 480 MHz) as the core, carrying AD7606 16-bit AD converter (sampling rate 200 kSPS); environmental temperature sensor DS18B20 (accuracy ±0.5℃), installed in the sunshade near the air conditioner outdoor unit; refrigerant high pressure sensor MPX5700 (range 0-4 MPa, accuracy ±1% FS), installed on the high pressure pipeline of the compressor exhaust port; vibration sensor ADXL357 (range 0-50g), installed on the top of the non-drive end bearing cover (vertical direction).

[0039] Internet of Things module: Edge computing layer uses Xilinx Zynq UltraScale+MPSoC development board (model ZCU102), carrying Linux operating system, running vibration analysis and voiceprint processing program; network layer uses Huawei ME909s-8214G module (supports LTE Cat.4) and Semtech SX1278 LoRa module (spreading factor SF=12), dual-mode communication module integrated into the edge computing development board; cloud platform layer deployed in Alibaba Cloud ECS server (4 cores 8G memory), InfluxDB database version 2.0, digital twin model developed with Unity 3D, supporting Web access (through Chrome browser).

[0040] Multi-machine group control module: Industrial-grade PLC (Siemens S7-1200, model 1214C) as the group control host, connected with 20 compressors through 485 communication interface (Modbus-RTU protocol), the communication address of each compressor is 1-20; PLC and Internet of Things module are connected through Ethernet (TCP / IP protocol), real-time acquisition of COP value, load rate and other data of each unit.

[0041] System software and algorithm implementation Load-environment prediction model: BP neural network uses 3-layer structure (input layer 3 neurons: environmental temperature, refrigerant high pressure fluctuation, running frequency; 10 neurons in hidden layer, activation function Sigmoid; 1 neuron in output layer: load prediction value), trained by MATLAB R2022b, training data set is 1000 sets of running data (sampling interval 10 min) of the mall in summer (June-August) 2023, the root mean square error (RMSE) of the trained model is 2.8%, and the mean absolute error (MAE) is 2.1%.

[0042] Dynamic weight MPC algorithm: implemented in STM32H743 using C language, the weight coefficient of the objective function is adjusted by a fuzzy logic controller (input: load rate, output: ω1, ω2, ω3), the fuzzy rule base contains 27 rules (e.g. "low load rate → large ω3"); the quadratic programming solution uses the QR decomposition method, with a solving time ≤20 ms; the carrier frequency of the SVPWM module is 10 kHz, and the output is to Infineon IKCM15F60GA IGBT module (600V / 15A), which drives the motor to run.

[0043] Fault diagnosis and early warning: the CNN diagnosis model of the edge computing layer is deployed using the TensorFlow Lite framework (model size 2MB), the input is 6 characteristic parameters of the vibration signal, and the output is 4 types of faults (normal, outer ring wear, inner ring wear, ball damage), the test set accuracy is 98.5%; the LSTM model of the cloud platform layer inputs 1000 groups of data in the past 24h (sampling interval 86.4s), and outputs the fault probability in the future 72h (output 1 value every 12h), trained using the Adam optimizer, with 500 iterations, and the loss function converges to 0.02.

[0044] Multi-function energy efficiency balanced scheduling: the scheduling algorithm in the PLC is implemented using ladder programming, which calculates the energy efficiency margin of each unit every 5s, and executes load distribution every 10s; when the system refrigeration demand decreases from 2200kW (20 units at full load) to 1100kW, the algorithm preferentially reduces the load of the 3 units with the lowest energy efficiency margin (operating for 3 years, actual COP 3.8), until they are shut down, and the load rate of the remaining 17 units is maintained at about 65%, the overall COP value increases from 3.9 to 4.3.

[0045] System testing and performance comparison This embodiment was tested in a large shopping mall (building area 50,000㎡) for 3 months (June-August 2024), and compared with the original traditional inductive motor compressor system (same power 110kW, total 20 units) of the mall, the test results are as follows: Energy efficiency comparison: the daily average energy consumption of the system is 820kWh, and that of the traditional system is 1180kWh, with a daily energy saving of 360kWh and an energy saving rate of 30.5%; during the summer peak period (14:00-16:00), the COP value of the system is 4.6, and that of the traditional system is 3.2, with a COP increase of 43.7%.

[0046] Response speed comparison: when the ambient temperature suddenly rises from 32℃ to 38℃ (load rate rises from 70% to 95%), the response time of the system of the application is 450ms (from detecting load change to adjusting to target frequency), and that of the traditional system is 1200ms, the response speed is increased by 62.5%; in terms of start-stop time, the start-up time of the system of the application is 1.2s (from start-up instruction to rated speed), and that of the traditional system is 2.8s, the start-up time is shortened by 57.1%, and the impact current is reduced from 800A to 320A (60%).

[0047] Fault early warning and maintenance comparison: during the test, the system of the application early warned the bearing outer ring wear failure of two units 48h in advance through the LSTM model, the maintenance personnel replaced the bearing in time, and no unexpected shutdown occurred; the traditional system had 3 unexpected shutdowns (all bearing failures) in the same period, each shutdown lasted for 48h, and the cumulative loss of refrigeration capacity was 21120kWh; the fault early warning accuracy of the system of the application is 96.3%, and the fault detection rate of the traditional system is only 65%.

[0048] Noise and vibration comparison: the running noise of the system of the application is 71dB (measured 1m away from the unit), and that of the traditional system is 78dB, the noise is reduced by 9%; the vibration acceleration of the motor bearing end cover is 0.22g, and that of the traditional system is 0.85g, the vibration is reduced by 74.1%.

[0049] Example 2: Screw-type intelligent refrigerant compressor system for cold chain logistics This embodiment is aimed at the cold chain logistics scene (such as large refrigeration warehouses), the characteristics of this scene are low refrigeration temperature (warehouse temperature-25℃-0℃), stable but long duration load (24h continuous operation), sensitive to refrigerant leakage (leakage will cause the temperature in the warehouse to rise, affecting the quality of goods), high requirements for the low-temperature adaptability, leakage detection capability and long-term reliability of the compressor, and the specific implementation details are as follows: System hardware configuration Permanent magnet synchronous motor module: a 160kW screw-type compressor dedicated permanent magnet synchronous motor is used, with a rated speed of 1800rpm, a peak torque of 450N・m, a speed regulation range of 500-3500rpm, a q-axis inductance to d-axis inductance ratio of 2.5; the rotor magnetic steel is N42SH type NdFeB (better low-temperature performance, magnetic flux density decreases ≤2% at-40℃), with an Al2O3 nano coating thickness of 6μm; the oil cooling channel diameter of the double cooling system is 10mm, the pitch is 25mm, the oil pump model is CB-B125 (maximum flow 125L / min), and the air cooling fan is a 140mm axial flow fan (model AFB1412VH, low-temperature start-up performance is good); the PT100 sensor of the magnetic steel health monitoring unit is a low-temperature type (-50℃-200℃, accuracy ±0.1℃), and the magnetic flux density detection coil output voltage range is 0-10V (corresponding to magnetic flux density 0-20T).

[0050] Intelligent control module: TI TMS320F28379D DSP (200 MHz) as the core, with ADS125624-bit AD converter (sampling rate 30kSPS, suitable for low temperature environment); environmental temperature sensor SHT30 (accuracy ±0.3℃, humidity measurement range 0-100%RH), installed in the ventilation of the cold storage outdoor unit; refrigerant high pressure sensor SCP1000 (range 0-6MPa, accuracy ±0.5%FS), installed in the compressor exhaust port; refrigerant low pressure sensor SCP0050 (range 0-0.5MPa, accuracy ±0.5%FS), installed in the compressor suction port; vibration sensor ADXL355 (range 0-20g, low temperature drift), installed in the horizontal direction of the motor drive end bearing cover.

[0051] Internet of Things module: Edge computing layer uses NVIDIA Jetson Nano development board (with Quad-core ARMA57 CPU), runs low-temperature adapted Linux system (kernel version 5.4), integrates vibration analysis, voiceprint collection and preliminary judgment of leakage functions; network layer uses ZTE MC801A 5G module (supports NRNSA / SA, transmission rate ≥100Mbps) and Semtech SX1262 LoRa module (spreading factor SF=10, communication distance 3km), dual-mode communication module has low-temperature start function (-40℃ can start normally); Cloud platform layer is deployed in Tencent Cloud CVM server (8 cores 16G memory), InfluxDB database uses cluster deployment (3 nodes), ensures data reliability, digital twin model supports mobile APP access (iOS / Android).

[0052] Multi-machine group control module: uses Schneider M258 PLC (model TM258LF42DT) as the group control host, connected with 30 compressors through Ethernet (Modbus-TCP protocol), the IP address of each compressor is 192.168.1.1-192.168.1.30; PLC is linked with the cold storage temperature control system (model Danfoss AK-CC550), adjusts the compressor load according to the temperature requirement in the warehouse (set at -18℃, fluctuation range ±1℃).

[0053] System software and algorithm optimization Low-temperature load prediction model: Considering the stable load in cold-chain scenarios but influenced by the opening and closing of the warehouse door (e.g., when loading and unloading, the door is opened, and the load rate temporarily increases by 10-15%), the input parameters of the BP neural network are increased by "the number of door opening and closing (collected by infrared sensors)", with 4 neurons in the input layer and 12 neurons in the hidden layer. The training data set is 1200 groups of data (including door opening and closing records) of a cold storage in winter (December-February) of 2023. After training, the prediction error of the model for temporary load is ≤2.5% — for example, when the door is opened and closed 5 times per hour, the model can predict that the load rate will increase from 80% to 92% in 5 minutes in advance.

[0054] Low-temperature optimization of dynamic weight MPC: In a low-temperature environment (motor housing temperature -10℃-5℃), the motor iron loss increases, so the fuzzy logic rules are adjusted: under the same load rate, ω3 increases by 0.1 in low temperature (preferably reduce energy consumption); at the same time, the objective function of MPC adds "winding temperature constraint term" ( , ), to avoid the insulation performance degradation caused by too low winding temperature (minimum temperature ≥-5℃).

[0055] Refrigerant leakage tracing optimization: To solve the problem of long pipeline (mostly 50-100m) and difficult positioning of leakage points in cold storage, the leakage tracing model in the cloud platform layer adds "pipeline length and sound speed correction term" — according to the pipeline material (stainless steel, sound speed 5800m / s) and length, the soundprint propagation time is corrected, and the positioning accuracy is improved from ±0.5m to ±0.3m; at the same time, an auxiliary microphone is installed every 10m in the pipeline (a total of 10), and through multi-microphone soundprint comparison, the tracing accuracy is further improved, and when the leakage rate is ≥0.05kg / h, it can be detected.

[0056] Low-temperature energy efficiency balancing of multi-machine group control: Considering that the energy efficiency difference of units is larger in low-temperature environment (new machine COP=3.8 at-25℃, old machine COP=3.2 after running for 2 years), the energy efficiency margin formula adds "temperature correction coefficient": , where T is the ambient temperature, the lower the ambient temperature, the smaller the correction coefficient, to avoid overloading of low-energy-efficiency units in low temperature — for example, when the ambient temperature is-20℃, the temperature correction coefficient=0.75, the energy efficiency margin of unit A (actual COP=3.5, load rate 70%) =0.75×(3.5 / 3.8)×0.3≈0.21, and the energy efficiency margin of unit B (actual COP=3.2, load rate 60%) =0.75×(3.2 / 3.8)×0.4≈0.25, preferentially allocate load to unit B.

[0057] System testing and performance verification The present embodiment is tested in a large cold chain logistics park (cold storage volume 50,000 m³, set temperature -18℃) for 6 months (January-June 2024), compared with the original 160 kW screw compressor system (traditional induction motor, total 30 units) of the logistics park, and the test results are as follows: Energy efficiency and operating cost comparison: the daily average energy consumption of the system is 1450 kWh, the traditional system is 2080 kWh, the daily energy saving is 630 kWh, the energy saving rate is 30.3%; according to the industrial electricity price of 0.8 yuan / kWh, the monthly electricity bill is saved by 15120 yuan, and the annual electricity bill is saved by 181,440 yuan; under low temperature working condition (temperature in the warehouse -25℃), the COP value of the system is 3.6, and the COP value of the traditional system is 2.5, and the COP value is increased by 44%.

[0058] Refrigerant leakage detection comparison: During the test, artificial simulation of refrigerant leakage was performed 3 times (leakage amount was 0.05 kg / h, 0.1 kg / h and 0.2 kg / h respectively), the detection time of the system was 8s, 5s and 3s respectively, and the positioning accuracy was ±0.3m, ±0.25m and ±0.2m respectively; the traditional system can only be found by pressure drop (the pressure drops from 0.8MPa to 0.6MPa after 24 hours of leakage), the detection time is ≥24h, and the leakage point cannot be located; the annual refrigerant loss of the system is reduced from 500kg of the traditional system to 50kg, which is reduced by 90%.

[0059] Low temperature reliability comparison: During the test, 3 cold waves (environmental temperature minimum -25℃) were experienced, the system had no shutdown, and the lowest temperature of the motor winding was -3℃ (meeting the insulation requirements); the traditional system had 2 shutdowns (both were motor failure to start, due to low temperature leading to increased lubricating oil viscosity), each shutdown time was 24h, which caused the warehouse temperature to rise to -10℃, and the loss of goods value was about 50,000 yuan; the average trouble-free working time (MTBF) of the system is 18000h, and that of the traditional system is 8000h, the reliability is increased by 125%.

[0060] Group control performance comparison: when 30 units are networked, the group control instruction transmission delay of the system is 350ms, the load distribution completion time is 8s, and the load rate deviation of each unit is ≤5%; the instruction delay of the traditional group control system is 1200ms, the load distribution completion time is 30s, and the load rate deviation of each unit is ≤15%; the overall energy efficiency of the system is increased by 22% compared with the traditional group control.

[0061] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. An intelligent refrigerant compressor system based on permanent magnet synchronous motor and Internet of Things, characterized in that, The permanent magnet synchronous motor module, the intelligent control module, the Internet of Things module, and the multi-machine group control module are sequentially electrically connected and data-interacted. The permanent magnet synchronous motor module includes a rotor, a stator, a double cooling system, and a magnetic steel health monitoring unit. The rotor adopts a Halbach array and a radial segmented hybrid magnetic circuit. The surface of the magnetic steel is covered with an Al2O3 nano anti-corrosion coating. The double cooling system includes an oil cooling channel arranged in a spiral shape inside the stator core and a wind cooling auxiliary device arranged on the motor end cover and including a temperature self-adaptive fan. The magnetic steel health monitoring unit includes a PT100 temperature sensor and a magnetic flux density detection coil. The intelligent control module includes a load-environment prediction unit, a dynamic weight MPC algorithm unit, and a vibration suppression unit. The load-environment prediction unit predicts load changes based on the data of the environmental temperature sensor and the refrigerant high-pressure sensor. The dynamic weight MPC algorithm unit adjusts the weight coefficients of torque, flux linkage, and energy consumption through fuzzy logic. The vibration suppression unit adopts FFT analysis and active damping algorithm. The Internet of Things module includes an edge computing layer, a network layer, and a cloud platform layer. The multi-machine group control module is based on an energy efficiency balanced scheduling algorithm and allocates loads according to the real-time COP values of each unit to avoid overloading of a single unit.

2. The intelligent refrigerant compressor system based on permanent magnet synchronous motor and Internet of Things according to claim 1, characterized in that: The stator of the permanent magnet synchronous motor module adopts a 12-slot 10-pole slot-pole matching, the ratio of q-axis inductance to d-axis inductance is greater than or equal to 2.2, the rated power is 50-315 kW, the speed regulation range is 500-3500 rpm, and the peak torque is 200-2000 N·m. 3.The intelligent refrigerant compressor system based on permanent magnet synchronous motor and Internet of Things according to claim 1, characterized in that: The load-environment prediction unit of the intelligent control module adopts a BP neural network prediction model. The input parameters include the average value of the environmental temperature in the past 10 minutes, the refrigerant high-pressure fluctuation value, and the compressor operating frequency. The output is the load prediction value in the next 5 minutes, and the prediction error is less than or equal to 3%.

4. The intelligent refrigerant compressor system based on permanent magnet synchronous motor and Internet of Things according to claim 1, characterized in that: The edge computing layer of the Internet of Things module also integrates a multi-scale entropy analysis unit for processing bearing vibration data collected by vibration sensors. Combined with a CNN diagnosis model, the bearing fault recognition rate is greater than 98%.

5. The intelligent refrigerant compressor system based on permanent magnet synchronous motor and internet of things of claim 1, wherein: The energy efficiency balanced scheduling algorithm of the multi-machine group control module includes the following steps: ① Collect the real-time COP values, load rates, and winding temperatures of each unit. ② Calculate the energy efficiency margins of each unit. ③ Distribute the new load to the unit with the highest energy efficiency margin until its load rate is less than or equal to 95%.

6. The intelligent refrigerant compressor system based on permanent magnet synchronous motor and internet of things of claim 1, wherein: The oil cooling channel of the double cooling system adopts a topology structure optimized by ANSYS Fluent. The oil flow is dynamically adjusted according to the stator temperature. When the temperature is greater than or equal to 120℃, the oil flow increases by 30%. The fan speed of the wind cooling auxiliary device is positively correlated with the temperature of the motor end cover. When the temperature increases by 10℃, the speed increases by 15%.

7. The intelligent refrigerant compressor system based on permanent magnet synchronous motor and internet of things of claim 1, wherein: The digital twin model of the cloud platform layer is based on an InfluxDB time series database. The data update period is less than or equal to 3 seconds. The life prediction adopts a Wiener process degradation modeling, and the prediction accuracy is greater than 88%. 8.The intelligent refrigerant compressor system based on PMSM and IoT of claim 1, wherein: The output voltage signal of the magnetic flux density detection coil of the magnetic steel health monitoring unit is transmitted to the intelligent control module after AD conversion. When the detected magnetic flux density decreases by more than 5%, the magnetic steel protection mode is triggered, and the motor output torque is reduced by 10%-20%.

9. The intelligent refrigerant compressor control method based on permanent magnet synchronous motor and Internet of Things, the method is implemented by using the intelligent refrigerant compressor control system based on permanent magnet synchronous motor and Internet of Things in any one of claims 1-8, characterized in that, The method includes the following steps: ①Initialization: The permanent magnet synchronous motor module is powered on, the Internet of Things module establishes edge-cloud communication, and the multi-machine group control module reads the initial state of each machine group; ②Load prediction: The intelligent control module collects environmental temperature and refrigerant pressure data, and outputs a 5-minute load prediction value through a BP neural network; ③Mode switching: The dynamic weight MPC algorithm unit adjusts the weight coefficient according to the load prediction value and switches to the economic / standard / overload mode; ④Real-time regulation: The vibration suppression unit analyzes the motor vibration frequency through FFT and outputs an active damping control signal; the dual cooling system adjusts the oil flow and fan speed according to the stator temperature and end cover temperature; ⑤Fault early warning and leakage tracing: The edge computing layer analyzes vibration and magnetic flux data, and the cloud platform layer uses an LSTM model to early warn of faults 72 hours in advance and locates the refrigerant leakage point in combination with voiceprint data; ⑥Multi-machine scheduling: The multi-machine group control module allocates loads according to the energy efficiency margin of each machine group to achieve optimal overall energy efficiency. 10.The intelligent refrigerant compressor control method based on a permanent magnet synchronous motor and an Internet of Things according to claim 9, characterized in that: The FFT analysis frequency range of the vibration suppression unit in step ④ is 0-2 kHz, and the response time of active damping control is ≤10 ms; In step ⑤, the refrigerant leakage tracing uses voiceprint feature matching, the feature extraction includes sound pressure level and frequency spectrum peak, and the positioning time is ≤10 s.