Defrosting control method for parallel flow evaporator

Abstract

The invention discloses a control method for defrosting of a parallel flow evaporator, which comprises the following steps of: constructing a multi-dimensional state sensing network, constructing a three-dimensional state sensing matrix by utilizing a distributed temperature sensor array on the surface of an evaporator coil, an air inlet humidity sensor and an inlet and outlet differential pressure type pressure sensor, and acquiring information such as temperature distribution, environment humidity and air pressure difference; through dynamic threshold self-adaptive judgment, based on an environment temperature field, the historical temperature of an evaporator and an icing critical temperature field calculated by a time sequence prediction model, the defrosting triggering condition is judged by combining humidity and an air pressure difference change gradient field; gradient type energy injection defrosting is adopted, in the first-stage defrosting process, the refrigerant flow is adjusted through a proportional electromagnetic valve to conduct directional defrosting, and in the second-stage defrosting process, a compressor is closed, a fin gap shape memory alloy heating network is activated, and latent heat of a phase change material is released to conduct uniform defrosting. And performing multi-physical field cooperative regulation and control, constructing a coupling model, and adjusting the heating power and pulse modulation duty ratio sequence.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of air conditioner control, in particular to a control method for defrosting of a parallel flow evaporator. BACKGROUND

[0002] As a key heat exchange component in air conditioning systems, the heat exchange efficiency of the parallel flow evaporator directly affects the overall operation performance and energy efficiency of the air conditioning equipment. In a low-temperature and high-humidity environment, frosting is prone to occur on the surface of the evaporator, which significantly increases the thermal resistance and reduces the heat exchange efficiency. Therefore, effective defrosting control of the parallel flow evaporator is an important link to ensure the stable and efficient operation of the air conditioning system.

[0003] In the prior art, the defrosting control of the parallel flow evaporator mainly relies on simple time triggering or single sensor monitoring, which is difficult to comprehensively obtain key information such as temperature distribution, humidity condition and air pressure change of the evaporator, resulting in inaccurate judgment of the frosting state, and often causing the defrosting opportunity to be too early or too late. The defrosting process mainly adopts a single mode of energy input, lacks a phased and targeted processing method, and is not only low in defrosting efficiency, but also easy to cause energy waste and difficult to realize uniform melting of the frost layer, which is prone to local overheating or residual frost layer. At the same time, the coordinated regulation of multiple physical fields such as temperature field, pressure field and frost thickness field is insufficient, which is difficult to avoid the anisotropy problem in the melting process of the frost layer, affecting the stable operation of the evaporator. The monitoring means of the defrosting process is limited, which is difficult to grasp the defrosting state in real time, and is prone to over-defrosting or incomplete defrosting, and lacks effective abnormal response mechanism, which is not conducive to the safety of the equipment. In addition, the existing system lacks continuous self-optimization ability, and is difficult to adaptively adjust the defrosting strategy according to different working conditions. Therefore, the defrosting effect and operation efficiency are easily affected in long-term use, and the adaptability and stability need to be improved. Therefore, a control method for defrosting of a parallel flow evaporator is proposed. SUMMARY

[0004] The present application provides the following technical scheme: a control method for defrosting of a parallel flow evaporator, comprising the following steps: S1, multi-dimensional state perception network construction: The temperature distribution is collected by the distributed temperature sensor array on the surface of the evaporator coil, the environmental humidity is obtained by the inlet humidity sensor, and the air pressure difference is monitored by the inlet and outlet differential pressure type pressure sensor, so as to construct a three-dimensional state perception matrix; S2, dynamic threshold self-adaptive judgment: Based on the environmental temperature field and the historical temperature of the evaporator, the icing critical temperature field is calculated by a time series prediction model. When the monitoring data deviates from the prediction model by more than a confidence interval and the humidity exceeds a dynamic threshold, the defrosting trigger condition is judged in combination with the air pressure difference change gradient field; S3, gradient type energy injection defrosting: Satisfy the conditions to start the first defrosting: open proportional electromagnetic valve, determine the opening curve according to the temperature and humidity field coupling model, adjust the refrigerant flow for directional defrosting; after the first defrosting, if the temperature of the key monitoring point is still lower than the dynamic threshold, start the second defrosting: close the compressor, activate the fin gap shape memory alloy heating network, and realize uniform defrosting through the latent heat release of the phase change material; S4, multi-physical field coordinated regulation: During heating defrosting, a temperature field-pressure field-frost layer thickness field coupling model is established, the heating power is adjusted by adaptive fuzzy control, and the pulse modulation duty cycle sequence is optimized based on the frost layer thickness field to ensure anisotropic melting of the frost layer and prevent local overheating; S5, intelligent monitoring of defrosting process: The surface temperature gradient and air pressure fluctuation are continuously monitored, and when the temperature reaches the phase change critical point and the air pressure fluctuation is normal, the defrosting is ended; if temperature changes or air pressure abnormalities occur, a three-level protection program is triggered, a redundant safety mechanism is started, and an abnormal feature map is generated; S6, closed-loop optimization decision system: After defrosting, the heating network is turned off, the compressor is restarted to restore the refrigeration cycle; a defrosting digital twin model is established, the current data is compared with the model, and the subsequent strategy is optimized through transfer learning, including adjusting the icing critical temperature field distribution, humidity threshold surface, and heating power density function.

[0005] The temperature distribution is collected by a distributed temperature sensor array on the surface of the evaporator coil, the environmental humidity is obtained by an inlet humidity sensor, and the air pressure difference is monitored by an inlet and outlet differential pressure type pressure sensor, thereby constructing a three-dimensional state perception matrix; Based on the environmental temperature field and the evaporator historical temperature, the icing critical temperature field is calculated using a time series prediction model. When the deviation of the monitoring data from the prediction model exceeds the confidence interval and the humidity exceeds the dynamic threshold, the defrosting trigger condition is judged in combination with the air pressure difference change gradient field. When the defrosting condition is met, the first defrosting is started: the proportional electromagnetic valve is opened, the opening curve is determined according to the temperature and humidity field coupling model, and the refrigerant flow is adjusted for directional defrosting; after the first defrosting, if the temperature of the key monitoring point is still lower than the dynamic threshold, the second defrosting is started: the compressor is closed, the fin gap shape memory alloy heating network is activated, and uniform defrosting is realized through the latent heat release of the phase change material; During the heating defrosting process, a temperature field-pressure field-frost layer thickness field coupling model is constructed, the heating power is adjusted by adaptive fuzzy control, and the pulse modulation duty cycle sequence is optimized based on the frost layer thickness field to ensure anisotropic melting of the frost layer and prevent local overheating; The evaporator surface temperature gradient and air pressure fluctuation are continuously monitored, and when the temperature reaches the phase change critical point and the air pressure fluctuation is normal, the defrosting is ended; if temperature changes or air pressure abnormalities occur, a three-level protection program is triggered, a redundant safety mechanism is started, and an abnormal feature map is generated; After the defrosting is completed, the heating network is closed, the compressor is restarted to resume the refrigeration cycle, a defrosting digital twin model is constructed, the defrosting data is compared with the model, and the subsequent defrosting strategy is optimized through transfer learning, including adjusting the icing critical temperature field distribution, the humidity threshold surface and the heating power density function.

[0006] Preferably, the distributed temperature sensor array uses a fiber grating temperature sensor, the humidity sensor network uses a graphene-based humidity sensor, and the differential pressure type pressure sensor uses a micro-electro-mechanical system pressure sensor.

[0007] In the system sensor configuration stage, the distributed temperature sensor array uses a fiber grating temperature sensor for deployment, the humidity sensor network uses a graphene-based humidity sensor for arrangement, and the differential pressure type pressure sensor uses a micro-electro-mechanical system pressure sensor for installation.

[0008] Preferably, in the S2, dynamic threshold adaptive determination, the combination of the air pressure difference change gradient field includes: when the air pressure difference change gradient exceeds the dynamically calculated critical gradient surface and the duration exceeds the characteristic time window, it is determined that the defrosting trigger condition is reached.

[0009] In the S2, dynamic threshold adaptive determination process, the air pressure difference change gradient field is monitored in real time, and when it is detected that the air pressure difference change gradient exceeds the dynamically calculated critical gradient surface and the duration exceeds the characteristic time window, the system determines that the defrosting trigger condition is reached.

[0010] Preferably, the proportional electromagnetic valve uses a magnetostrictive drive mechanism to achieve nanoscale precise control of the valve opening degree through the strain effect of the super-magnetostrictive material.

[0011] When the proportional electromagnetic valve is running, a magnetostrictive drive mechanism is used to achieve nanoscale precise control of the valve opening degree through the strain effect of the super-magnetostrictive material by controlling the strain of the material.

[0012] Preferably, in the S3, gradient energy injection defrosting, the activation mode of the shape memory alloy heating network includes two modes of electric pulse excitation and thermal conduction activation, and the two modes can work independently or cooperatively, wherein the electric pulse excitation mode adjusts the heating power by controlling the pulse frequency and duty cycle, and the thermal conduction activation mode adjusts the heating uniformity by adjusting the heat diffusion rate of the phase change material.

[0013] In the S3, gradient energy injection defrosting, the activation of the shape memory alloy heating network uses two modes of electric pulse excitation and thermal conduction activation, and the two modes can work independently or cooperatively; the electric pulse excitation mode changes the heating power by adjusting the pulse frequency and duty cycle, and the thermal conduction activation mode adjusts the heating uniformity by adjusting the heat diffusion rate of the phase change material.

[0014] Preferably, in the S4, multi-physical field synergistic regulation, further comprising: A frost melting prediction model is established by a Bayesian optimization algorithm to update the duty cycle sequence of the pulse width modulation in real time.

[0015] In the S4, multi-physical field synergistic regulation, a Bayesian optimization algorithm is introduced to construct a frost melting prediction model, which is operated according to real-time data to update the duty cycle sequence of the pulse width modulation in real time.

[0016] Preferably, in the S5, intelligent monitoring of defrosting process, further comprising: The temperature field gradient data is subjected to multi-scale decomposition by wavelet transform to extract the characteristic frequency of frost melting.

[0017] In the S5, intelligent monitoring of defrosting process, the temperature field gradient data obtained is subjected to multi-scale decomposition by wavelet transform to extract the characteristic frequency of frost melting.

[0018] Preferably, after the start of the three-level protection program, the system automatically generates a fault diagnosis report containing an abnormal feature map and sends it to the edge computing node through a low-power wide-area network for deep analysis.

[0019] After the start of the three-level protection program, the system automatically analyzes the abnormal situation, generates a fault diagnosis report containing an abnormal feature map, and then sends the report to the edge computing node through a low-power wide-area network for deep analysis by the edge computing node.

[0020] Preferably, in the S6, closed-loop optimization decision system, further comprising: The historical defrosting data is subjected to knowledge graph construction to extract optimization rules of defrosting strategies.

[0021] In the S6, closed-loop optimization decision system, historical defrosting data is collected and subjected to knowledge graph construction to extract optimization rules of defrosting strategies through the knowledge graph.

[0022] Preferably, the transfer learning adopts a federated learning architecture to realize global optimization of defrosting strategies through collaborative training of edge nodes and cloud servers.

[0023] In the transfer learning process, a federated learning architecture is adopted for collaborative training of edge nodes and cloud servers to realize global optimization of defrosting strategies through information interaction and joint operation of the two.

[0024] In summary, compared with the prior art, the present application provides a control method for defrosting of a parallel flow evaporator, which has the following advantages: The application can comprehensively and accurately obtain the temperature distribution, environmental humidity and air pressure difference and other state information of the evaporator by constructing a multi-dimensional state perception network, using a three-dimensional state perception matrix constructed by a distributed temperature sensor array on the surface of the evaporator coil, an inlet air humidity sensor and an inlet and outlet differential pressure sensor, to provide reliable data support for subsequent defrosting control. By dynamic threshold adaptive determination, the defrosting trigger condition is judged based on the icing critical temperature field calculated by the environmental temperature field, the evaporator historical temperature and the time sequence prediction model, combined with the humidity and air pressure difference gradient field, so that the defrosting opportunity can be accurately controlled, and energy waste or heat exchange efficiency reduction caused by defrosting too early or too late can be avoided. Gradient energy injection defrosting is adopted, the first-stage defrosting is directional defrosting by adjusting the refrigerant flow according to the temperature and humidity field coupling model through a proportional electromagnetic valve, the second-stage defrosting is uniform defrosting by closing the compressor and activating the fin gap shape memory alloy heating network and the latent heat release of the phase change material, so that the defrosting operation can be carried out in stages and targeted, the defrosting efficiency is improved, and the energy consumption is reduced. Multi-physical field cooperative regulation and control is achieved by constructing a temperature field-pressure field-frost layer thickness field coupling model, adjusting the heating power by using adaptive fuzzy control and optimizing the pulse modulation duty cycle sequence based on the frost layer thickness field, so that the anisotropic melting of the frost layer can be ensured, local overheating can be effectively prevented, and the stable operation of the evaporator can be ensured. Intelligent monitoring of defrosting process is achieved by continuously monitoring the surface temperature gradient and air pressure fluctuation, ending defrosting when the temperature reaches the phase change critical point and the air pressure fluctuation is normal, triggering a three-stage protection program and generating an abnormal feature map when the temperature reaches the phase change critical point and the air pressure fluctuation is abnormal, so that the defrosting process can be monitored in real time, defrosting can be terminated in time to avoid over-defrosting, and the safety of the equipment can be ensured when an abnormality occurs. The closed-loop optimization decision system is achieved by constructing a defrosting digital twin model, comparing the current defrosting data with the model and optimizing the subsequent strategy through transfer learning, including adjusting the icing critical temperature field distribution, the humidity threshold surface and the heating power density function, so that the system can continuously adapt to different working conditions, continuously improve the defrosting effect and operation efficiency, and enhance the adaptability and stability of the parallel flow evaporator in the long-term use process. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 is a schematic block diagram of the control method of the application. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0027] See also Figure 1 The present invention provides a technical solution, a control method for defrosting a parallel flow evaporator, comprising the following steps: S1. Construction of multi-dimensional state perception network: The distributed temperature sensor array on the evaporator coil surface collects temperature distribution, the air inlet humidity sensor obtains ambient humidity, and the inlet and outlet differential pressure sensors monitor the air pressure difference to build a three-dimensional state perception matrix; S2, dynamic threshold adaptive judgment: Based on the ambient temperature field and the historical evaporator temperature, the time series prediction model calculates the critical freezing temperature field. When the deviation between the monitoring data and the prediction model exceeds the confidence interval and the humidity exceeds the dynamic threshold, the defrost trigger condition is determined by combining the pressure difference gradient field. S3, gradient energy injection defrost: If the conditions are met, the first-stage defrost is initiated: the proportional solenoid valve is opened, the opening curve is determined according to the temperature and humidity field coupling model, and the refrigerant flow is adjusted for directional defrosting. If the temperature of the key monitoring point is still below the dynamic threshold after the first stage, the second-stage defrost is initiated: the compressor is turned off, the shape memory alloy heating network in the fin gap is activated, and the latent heat of the phase change material is released to evenly defrost. S4. Multi-physics field coordinated control: During heating and defrosting, a coupled model of temperature field, pressure field, and frost thickness field is established. Adaptive fuzzy control is used to adjust the heating power. The pulse modulation duty cycle sequence is optimized based on the frost thickness field to ensure anisotropic melting of the frost layer and prevent local overheating. S5, intelligent monitoring of defrost process: Continuously monitor the surface temperature gradient and air pressure fluctuations. When the temperature reaches the critical point of phase change and the air pressure fluctuations are normal, the defrost ends. When temperature change or air pressure abnormality occurs, the third-level protection program is triggered, the redundant safety mechanism is activated, and an abnormal characteristic map is generated. S6, Closed-loop optimization decision system: After defrosting is completed, the heating network is turned off and the compressor is restarted to resume the refrigeration cycle; a defrosting digital twin model is established, the data is compared with the model, and transfer learning is used to optimize subsequent strategies, including adjusting the critical temperature field distribution of freezing, the humidity threshold surface and the heating power density function.

[0028] The temperature distribution is collected through a distributed temperature sensor array on the surface of the evaporator coil, the ambient humidity is obtained using an air inlet humidity sensor, and the air pressure difference is monitored using an inlet and outlet differential pressure sensor, thereby constructing a three-dimensional state perception matrix; Based on the ambient temperature field and the historical evaporator temperature, a time series prediction model is used to calculate the critical freezing temperature field. When the deviation between the monitoring data and the prediction model exceeds the confidence interval and the humidity exceeds the dynamic threshold, the defrost trigger condition is determined by combining the pressure difference gradient field. When the defrosting condition is met, the first-stage defrosting is started: the proportional electromagnetic valve is opened, the opening curve is determined according to the temperature and humidity field coupling model, and the refrigerant flow is adjusted for directional defrosting; after the first-stage defrosting, if the temperature of the key monitoring point is still lower than the dynamic threshold, the second-stage defrosting is started: the compressor is turned off, the fin gap shape memory alloy heating network is activated, and uniform defrosting is realized through latent heat release of the phase change material; During the heating defrosting process, a temperature field-pressure field-frost layer thickness field coupling model is constructed, the heating power is adjusted through adaptive fuzzy control, and the pulse modulation duty cycle sequence is optimized based on the frost layer thickness field, so as to ensure anisotropic melting of the frost layer and prevent local overheating; The surface temperature gradient and air pressure fluctuation of the evaporator are continuously monitored, and the defrosting is ended when the temperature reaches the phase change critical point and the air pressure fluctuation is normal; if temperature change or air pressure abnormality occurs, the third-stage protection program is triggered, the redundant safety mechanism is started, and an abnormal feature map is generated; After the defrosting is completed, the heating network is turned off, and the compressor is restarted to restore the refrigeration cycle; a defrosting digital twin model is constructed, the defrosting data of this time is compared with the model, and the subsequent defrosting strategy is optimized through transfer learning, including adjusting the icing critical temperature field distribution, the humidity threshold surface and the heating power density function; The multi-dimensional state perception network construction link collects temperature distribution, environmental humidity and air pressure difference through multiple sensors respectively and constructs a three-dimensional state perception matrix, which can comprehensively and accurately obtain the running state information of the evaporator, providing reliable basic data support for subsequent defrosting judgment and control; The dynamic threshold adaptive judgment link calculates the icing critical temperature field based on the environmental temperature field and the historical temperature of the evaporator, judges the defrosting trigger condition in combination with various monitoring data, makes the defrosting trigger judgment more accurate, can adapt to different environmental changes and evaporator running states, and avoids unnecessary defrosting operation or defrosting not in time; Different defrosting methods are used in different stages according to the defrosting situation, the first-stage defrosting is directional defrosting by adjusting the refrigerant flow, and the second-stage defrosting is uniform defrosting by using the shape memory alloy heating network and the latent heat release of the phase change material, which realizes gradient injection of energy, improves the pertinence and efficiency of defrosting, and reduces energy waste; The multi-physical field coupling model is constructed, the heating power is adjusted through adaptive fuzzy control, and the pulse modulation duty cycle sequence is optimized, which ensures anisotropic melting of the frost layer and prevents local overheating, guarantees the uniformity and safety of the defrosting process, and avoids damage to the evaporator caused by local overheating; The defrosting process is judged by continuously monitoring the surface temperature gradient and air pressure fluctuation, the defrosting is ended in time under normal circumstances, and the third-stage protection program is triggered when abnormality occurs, which improves the reliability and safety of the defrosting process and can effectively cope with various abnormal situations, By constructing a defrosting digital twin model, the defrosting data is used to optimize the subsequent defrosting strategy, realizing the continuous iteration and upgrading of the defrosting strategy, so that the defrosting method has self-learning and self-optimization capabilities, and can continuously improve the defrosting performance as the use time increases, adapting to various changes in the long-term use process.

[0029] The distributed temperature sensor array uses fiber grating temperature sensors, the humidity sensor network uses graphene-based humidity sensors, and the differential pressure pressure sensor uses micro-electro-mechanical system pressure sensors.

[0030] In the system sensor configuration phase, the distributed temperature sensor array uses fiber grating temperature sensors for deployment, the humidity sensor network uses graphene-based humidity sensors for arrangement, and the differential pressure pressure sensor uses micro-electro-mechanical system pressure sensors for installation. The fiber grating temperature sensor has good anti-electromagnetic interference ability and stability, and can accurately sense temperature distribution; the graphene-based humidity sensor has high sensitivity and can quickly respond to humidity changes; the micro-electro-mechanical system pressure sensor has small size and high precision, and can accurately measure pressure difference, which improves the accuracy and reliability of the system in sensing environmental parameters.

[0031] In the dynamic threshold adaptive determination S2, the pressure difference change gradient field is combined: when the pressure difference change gradient exceeds the dynamically calculated critical gradient surface and the duration exceeds the characteristic time window, it is determined that the defrosting trigger condition is reached.

[0032] In the dynamic threshold adaptive determination process S2, the pressure difference change gradient field is monitored in real time, and when it is detected that the pressure difference change gradient exceeds the dynamically calculated critical gradient surface and the state duration exceeds the characteristic time window, the system determines that the defrosting trigger condition is reached. The accuracy of the defrosting trigger condition determination is improved by combining the pressure difference change gradient field and the duration, which can avoid false triggering caused by instantaneous fluctuations, while ensuring that defrosting is started in time when it is really needed, and optimizing the control of defrosting opportunity.

[0033] The proportional electromagnetic valve uses a magnetostrictive drive mechanism to achieve nanoscale precise control of valve opening through the strain effect of super magnetostrictive materials.

[0034] When the proportional electromagnetic valve is running, a magnetostrictive drive mechanism is used to achieve nanoscale control through the strain effect of super magnetostrictive materials, which can precisely control the valve opening by controlling the strain of the material. Nanoscale control is achieved through the strain effect of super magnetostrictive materials, which improves the precision of proportional electromagnetic valve opening control and allows more precise regulation of fluid flow, enhancing the precision and stability of the system in controlling related parameters.

[0035] S3, in the gradient energy injection defrosting, the activation mode of the shape memory alloy heating network includes two modes of electric pulse excitation and heat conduction activation, and the two modes can work independently or cooperatively, wherein the electric pulse excitation mode adjusts the heating power by controlling the pulse frequency and duty cycle, and the heat conduction activation mode adjusts the heating uniformity by adjusting the heat diffusion rate of the phase change material.

[0036] In S3, gradient energy injection defrosting, the activation of the shape memory alloy heating network adopts two modes of electric pulse excitation and heat conduction activation, and the two modes can be operated independently or cooperatively; the electric pulse excitation mode changes the heating power by adjusting the pulse frequency and duty cycle, and the heat conduction activation mode adjusts the heating uniformity by adjusting the heat diffusion rate of the phase change material; The independent or cooperative work of the two activation modes increases the flexibility of the heating mode, which can adapt to different defrosting requirements; by adjusting the pulse frequency and duty cycle, the heating power can be accurately controlled, and by adjusting the heat diffusion rate of the phase change material, the heating uniformity can be improved, thereby improving the defrosting efficiency and effect.

[0037] S4, multi-physical field cooperative regulation, further includes: A frost melting prediction model is established by using Bayesian optimization algorithm to update the duty cycle sequence of pulse width modulation in real time.

[0038] In S4, multi-physical field cooperative regulation, Bayesian optimization algorithm is introduced, and a frost melting prediction model is constructed using the algorithm. The model operates according to real-time data to update the duty cycle sequence of pulse width modulation in real time. The Bayesian optimization algorithm can efficiently construct the prediction model, and the real-time updated duty cycle sequence can make the pulse width modulation more suitable for the actual situation of frost melting, thereby improving the accuracy of energy injection during defrosting and optimizing the defrosting effect.

[0039] S5, intelligent monitoring of defrosting process, further includes: Wavelet transform is used to perform multi-scale decomposition on the temperature field gradient data to extract the characteristic frequency of frost melting.

[0040] In S5, intelligent monitoring of defrosting process, wavelet transform is used to perform multi-scale decomposition on the obtained temperature field gradient data to extract the characteristic frequency of frost melting. The multi-scale decomposition of wavelet transform can effectively process the temperature field gradient data, facilitate accurate extraction of the characteristic frequency of frost melting, and improve the sensitivity and accuracy of monitoring the defrosting process, which helps to timely grasp the frost melting state.

[0041] After the three-level protection program is started, the system automatically generates a fault diagnosis report containing an abnormal feature map and sends it to the edge computing node through a low-power wide-area network for deep analysis.

[0042] After the three-level protection program is started, the system automatically analyzes the abnormal situation, generates a fault diagnosis report containing an abnormal feature map, and then sends the report to the edge computing node through the low-power wide-area network for deep analysis by the edge computing node; The automatic generation of the fault diagnosis report and the sending to the edge node for analysis can quickly realize the deep analysis of the abnormal situation, and the low-power wide-area network ensures the stability and energy saving of data transmission, thereby improving the timeliness and professionalism of system fault handling.

[0043] In the closed-loop optimization decision system S6, further comprising: The historical defrosting data is used to construct a knowledge graph to extract defrosting strategy optimization rules.

[0044] In the closed-loop optimization decision system S6, historical defrosting data is collected, and a knowledge graph is constructed for these data to extract optimization rules of defrosting strategies through the knowledge graph; Based on historical data, a knowledge graph is constructed and optimization rules are extracted, which can fully utilize past experience to guide defrosting strategy optimization, making defrosting decisions more scientific and targeted, and improving the intelligent level of the system.

[0045] The transfer learning adopts a federated learning architecture to realize global optimization of defrosting strategies through collaborative training of edge nodes and cloud servers.

[0046] In the transfer learning process, a federated learning architecture is adopted, and edge nodes and cloud servers are collaboratively trained to realize global optimization of defrosting strategies through information interaction and joint operation of the two; The collaborative training under the federated learning architecture realizes the complementary advantages of edge nodes and the cloud while protecting data privacy, which helps to form a more optimal global defrosting strategy and improve the applicability and effectiveness of the defrosting strategy.

[0047] In this scheme, the distributed temperature sensor array is deployed with fiber grating temperature sensors, which have good anti-electromagnetic interference ability and stability, and can accurately perceive temperature distribution; the humidity sensor network is arranged with graphene-based humidity sensors, which have high sensitivity and can quickly respond to humidity changes; the differential pressure type pressure sensor is installed with a micro-electromechanical system pressure sensor, which has small size and high precision, and can accurately measure pressure difference, and the three work together to improve the accuracy and reliability of the system in perceiving environmental parameters; The temperature distribution is collected through a distributed temperature sensor array on the surface of the evaporator coil; the ambient humidity is obtained using an air inlet humidity sensor; and the air pressure difference is monitored using an inlet and outlet differential pressure sensor. This constructs a three-dimensional state perception matrix that can comprehensively and accurately obtain information on the evaporator's operating status, providing reliable basic data support for subsequent defrost judgment and control. Based on the ambient temperature field and the historical temperature of the evaporator, the critical temperature field for freezing is calculated using a time series prediction model; the pressure difference change gradient field is monitored in real time; when the deviation between the monitoring data and the prediction model exceeds the confidence interval and the humidity exceeds the dynamic threshold, the defrost triggering condition is judged in combination with the pressure difference change gradient field; when it is detected that the pressure difference change gradient exceeds the dynamically calculated critical gradient surface, and the duration of this state exceeds the characteristic time window, the system determines that the defrost triggering condition is met, and the judgment is made in combination with the pressure difference change gradient field and the duration, which improves the accuracy of the defrost triggering condition judgment, avoids false triggering caused by instantaneous fluctuations, and ensures timely start when defrosting is really needed, optimizes the control of defrosting timing, makes the judgment of defrosting trigger more accurate, can adapt to different environmental changes and evaporator operating conditions, and avoids unnecessary defrosting operations or untimely defrosting; When the defrosting conditions are met, the first-level defrost is started: the proportional solenoid valve is opened. When the proportional solenoid valve is running, a magnetostrictive drive mechanism is used, and the strain effect of the giant magnetostrictive material is utilized to control the strain of the material to adjust the valve opening, so as to achieve nanometer-level precise control. With the help of the strain effect of the giant magnetostrictive material, nanometer-level control is achieved, which improves the accuracy of the proportional solenoid valve opening control, can more accurately adjust the fluid flow, and enhance the fineness and stability of the system's control over related parameters. The opening curve is determined according to the temperature and humidity field coupling model, and the refrigerant flow is adjusted for directional defrosting; after the first-level defrost, if the temperature of the key monitoring point is still lower than the dynamic threshold, the second-level defrost is started: the compressor is turned off, and the fin gap shape memory alloy heating network is activated. The shape memory alloy heating network The network is activated using two modes: electric pulse excitation and heat conduction activation. The two modes can operate independently or work together. The electric pulse excitation mode changes the heating power by adjusting the pulse frequency and duty cycle, and the heat conduction activation mode adjusts the heating uniformity by adjusting the thermal diffusion rate of the phase change material. The two activation modes work independently or in conjunction with each other, which increases the flexibility of the heating method and can adapt to different defrosting needs. The heating power can be precisely controlled by adjusting the pulse frequency and duty cycle, and the heating uniformity can be improved by adjusting the thermal diffusion rate of the phase change material, thereby improving the defrosting efficiency and effect. Uniform defrosting is achieved through the release of latent heat from the phase change material. Different defrosting methods are adopted in stages according to the defrosting situation, realizing gradient energy injection, improving the targetedness and efficiency of defrosting, and reducing energy waste. In the heating defrosting process, a temperature field-pressure field-frost layer thickness field coupling model is constructed, the heating power is adjusted through adaptive fuzzy control, and the pulse modulation duty cycle sequence is optimized based on the frost layer thickness field to ensure anisotropic melting of the frost layer and prevent local overheating; A Bayesian optimization algorithm is introduced, and a frost melting prediction model is constructed using the algorithm. The model operates based on real-time data to update the pulse width modulation duty cycle sequence in real time. The Bayesian optimization algorithm can efficiently construct the prediction model, and the real-time updated duty cycle sequence can make the pulse width modulation more suitable for the actual situation of frost melting, improving the precision of energy injection during the defrosting process and optimizing the defrosting effect. The construction of a multi-physical field coupling model and the adjustment of the heating power and optimization of the pulse modulation duty cycle sequence through adaptive fuzzy control ensure the uniformity and safety of the defrosting process and prevent damage to the evaporator due to local overheating; The temperature gradient and air pressure fluctuation of the evaporator surface are continuously monitored. When the temperature reaches the phase transition critical point and the air pressure fluctuation is normal, the defrosting process is ended. If there is a temperature change or abnormal air pressure, a three-level protection program is triggered, a redundant safety mechanism is started, and an abnormal feature map is generated. The temperature field gradient data obtained is decomposed by wavelet transform at multiple scales, and the characteristic frequency of frost melting is extracted through the decomposition process. The multi-scale decomposition of wavelet transform can effectively process the temperature field gradient data, facilitate the accurate extraction of the characteristic frequency of frost melting, and improve the sensitivity and accuracy of defrosting process monitoring, which helps to timely grasp the frost melting state. The defrosting process is judged by continuously monitoring the surface temperature gradient and air pressure fluctuation. In normal cases, the defrosting process is ended in time, and when an abnormality occurs, a three-level protection program is triggered, which improves the reliability and safety of the defrosting process and can effectively deal with various abnormal situations. After the three-level protection program is started, the system automatically analyzes the abnormal situation and generates a fault diagnosis report containing the abnormal feature map, which is then sent to the edge computing node through a low-power wide-area network for deep analysis by the edge computing node. The automatic generation of the fault diagnosis report and its transmission to the edge node for analysis can quickly realize the deep analysis of abnormal situations, and the low-power wide-area network ensures the stability and energy saving of data transmission, improving the timeliness and professionalism of system fault handling; After the defrosting is completed, the heating network is closed, the compressor is restarted to resume the refrigeration cycle; a defrosting digital twin model is constructed, the defrosting data of this time is compared with the model, and the subsequent defrosting strategy is optimized through transfer learning, including adjusting the icing critical temperature field distribution, the humidity threshold surface and the heating power density function; historical defrosting data is collected, and the knowledge graph is constructed through the knowledge graph, and the optimization rules of the defrosting strategy are extracted; based on the historical data, the knowledge graph is constructed and the optimization rules are extracted, which can fully utilize the past experience to guide the defrosting strategy optimization, so that the defrosting decision is more scientific and targeted, and the intelligent level of the system is improved; the transfer learning adopts a federated learning architecture, and the global optimization of the defrosting strategy is realized through the collaborative training of the edge node and the cloud server; the collaborative training under the federated learning architecture realizes the complementary advantages of the edge node and the cloud while ensuring data privacy, which helps to form a better global defrosting strategy, improves the applicability and effectiveness of the defrosting strategy, and through the construction of the defrosting digital twin model, the defrosting data of this time is used to optimize the subsequent defrosting strategy, realizing the continuous iteration and upgrading of the defrosting strategy, so that the defrosting method has the ability of self-learning and self-optimization, and can continuously improve the defrosting performance with the increase of the use time, adapting to various changes in the long-term use process.

[0048] It should be noted that, in the present document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0049] While embodiments of the present application have been shown and described with reference to particular embodiments thereof, it will be understood by those skilled in the art that various changes in form and details can be made therein without departing from the spirit and scope of the application. The scope of the application is defined by the appended claims and their equivalents.

Claims

1. A control method for defrosting a parallel flow evaporator, characterized in that: The steps include: S1. Construction of multi-dimensional state perception network: The distributed temperature sensor array on the evaporator coil surface collects temperature distribution, the air inlet humidity sensor obtains ambient humidity, and the inlet and outlet differential pressure sensors monitor the air pressure difference to build a three-dimensional state perception matrix; S2, dynamic threshold adaptive judgment: Based on the ambient temperature field and the historical evaporator temperature, the time series prediction model calculates the critical freezing temperature field. When the deviation between the monitoring data and the prediction model exceeds the confidence interval and the humidity exceeds the dynamic threshold, the defrost trigger condition is determined by combining the pressure difference gradient field. S3, gradient energy injection defrost: If the conditions are met, the first-stage defrost is initiated: the proportional solenoid valve is opened, the opening curve is determined according to the temperature and humidity field coupling model, and the refrigerant flow is adjusted for directional defrosting. If the temperature of the key monitoring point is still below the dynamic threshold after the first stage, the second-stage defrost is initiated: the compressor is turned off, the shape memory alloy heating network in the fin gap is activated, and the latent heat of the phase change material is released to evenly defrost. S4. Multi-physics field coordinated control: During heating and defrosting, a coupled model of temperature field, pressure field, and frost thickness field is established. Adaptive fuzzy control is used to adjust the heating power. The pulse modulation duty cycle sequence is optimized based on the frost thickness field to ensure anisotropic melting of the frost layer and prevent local overheating. S5, intelligent monitoring of defrost process: Continuously monitor the surface temperature gradient and air pressure fluctuations. When the temperature reaches the critical point of phase change and the air pressure fluctuations are normal, the defrost ends. When temperature change or air pressure abnormality occurs, the third-level protection program is triggered, the redundant safety mechanism is activated, and an abnormal characteristic map is generated. S6, Closed-loop optimization decision system: After defrosting is completed, the heating network is turned off and the compressor is restarted to resume the refrigeration cycle; a defrosting digital twin model is established, the data is compared with the model, and transfer learning is used to optimize subsequent strategies, including adjusting the critical temperature field distribution of freezing, the humidity threshold surface and the heating power density function.

2. The method for controlling defrosting of a parallel flow evaporator according to claim 1, characterized in that: The distributed temperature sensor array adopts a fiber Bragg grating temperature sensor, the humidity sensor network adopts a graphene-based humidity sensor, and the differential pressure sensor adopts a micro-electromechanical system pressure sensor.

3. The method for controlling defrosting of a parallel flow evaporator according to claim 1, characterized in that: In the dynamic threshold adaptive determination in S2, combining the pressure difference change gradient field includes: when the pressure difference change gradient exceeds the dynamically calculated critical gradient surface and the duration exceeds the characteristic time window, it is determined that the defrost trigger condition is met.

4. The method for controlling defrosting of a parallel flow evaporator according to claim 1, characterized in that: The proportional solenoid valve adopts a magnetostrictive driving mechanism and realizes nanometer-level precise control of valve opening through the strain effect of giant magnetostrictive material.

5. The method for controlling defrosting of a parallel flow evaporator according to claim 1, characterized in that: In the S3, gradient energy injection defrosting, the activation methods of the shape memory alloy heating network include two modes: electric pulse excitation and heat conduction activation, and the two modes work independently or in coordination. The electric pulse excitation mode adjusts the heating power by controlling the pulse frequency and duty cycle, and the heat conduction activation mode adjusts the heating uniformity by the thermal diffusion rate of the phase change material.

6. The method for controlling defrosting of a parallel flow evaporator according to claim 1, characterized in that: Said S4, multi-physics field coordinated control, further includes: A frost melting prediction model is established using the Bayesian optimization algorithm to update the duty cycle sequence of pulse width modulation in real time.

7. The method for controlling defrosting of a parallel flow evaporator according to claim 1, characterized in that: The above-mentioned S5, intelligent monitoring of the defrosting process, further includes: Wavelet transform is used to perform multi-scale decomposition of temperature field gradient data to extract the characteristic frequency of frost melting.

8. The method for controlling defrosting of a parallel flow evaporator according to claim 1, characterized in that: After the third-level protection program is started, the system automatically generates a fault diagnosis report containing an abnormal feature map and sends it to the edge computing node through a low-power wide area network for in-depth analysis.

9. The method for controlling defrosting of a parallel flow evaporator according to claim 1, characterized in that: Said S6, closed-loop optimization decision system, further comprising: A knowledge graph is constructed based on historical defrost data to extract defrost strategy optimization rules.

10. The method for controlling defrosting of a parallel flow evaporator according to claim 1, characterized in that: The transfer learning adopts a federated learning architecture to achieve global optimization of the defrosting strategy through collaborative training of edge nodes and cloud servers.

Images