Refrigeration unit capable of effectively utilizing a liquid accumulator and control method thereof

Abstract

This invention belongs to the field of refrigeration equipment, and particularly relates to a refrigeration unit and its control method that can effectively utilize a liquid receiver. The invention incorporates a pressure regulating pipe and a dedicated heat exchanger at the top of the liquid receiver, constructing a bidirectional heat exchange loop between the uncondensed gaseous refrigerant in the liquid receiver and the return gas from the cooling fan. This achieves secondary condensation of the uncondensed gas and superheating of the return gas, eliminating the risk of gas accumulation in the liquid receiver and liquid slugging in the compressor. In terms of the control method, an empirical wavelet transform filtering algorithm based on thermodynamic consistency constraints is proposed to improve the accuracy of the state signal. An improved dual-subgroup memory cooperative bat algorithm is used to optimize the parameters of the active disturbance rejection controller, achieving high-precision adjustment of the expansion valve. This invention has a simple structure, low modification cost, and can significantly improve the effective utilization rate of the liquid receiver and system energy efficiency, extend the service life of the equipment, and is applicable to various refrigeration systems such as industrial refrigeration and commercial cold chain systems.

Description

Technical Field

[0001] This invention belongs to the field of refrigeration equipment, and in particular relates to a refrigeration unit that can effectively utilize a liquid receiver and its control method. Background Technology

[0002] Refrigeration units are widely used in industrial refrigeration, commercial cold chain, air conditioning, and many other fields. The liquid receiver, as the core energy storage and pressure stabilizing component of the refrigeration system, is mainly used to store liquid refrigerant, balance the refrigerant circulation volume under different operating conditions, and stabilize system pressure and buffer pipeline flow fluctuations. It is a key structure to ensure the continuous and stable operation of the refrigeration unit. Under normal operating conditions, the high-pressure liquid refrigerant, after being fully condensed by the condenser, enters the liquid receiver for storage. The liquid receiver is mainly composed of liquid refrigerant, with only a small amount of pressure-stabilizing gas remaining. This ensures precise and controllable refrigerant circulation volume, guaranteeing heat exchange efficiency and operational stability.

[0003] Currently, conventional refrigeration units generally suffer from poor condensation heat exchange efficiency in actual operation, making it difficult to fully condense the high-temperature, high-pressure gaseous refrigerant discharged from the compressor into saturated liquid refrigerant. On the one hand, the operating conditions of refrigeration units fluctuate significantly. Under conditions such as high unit load, high ambient temperature, dust accumulation and blockage of condenser heat dissipation fins, or decreased fan cooling efficiency, the overall heat transfer coefficient of the condenser decreases drastically, and the heat dissipation capacity cannot match the condensation requirements of the gaseous refrigerant. A large amount of gaseous refrigerant cannot complete phase change liquefaction and flows directly into the liquid receiver after only a slight temperature drop. On the other hand, the existing condenser's piping structure and heat exchange layout are fixed and cannot adaptively adjust the condensation efficiency according to changes in system load and ambient temperature, easily leading to incomplete condensation and the mixing of gas and liquid phases.

[0004] The aforementioned defects in the existing technology directly lead to the accumulation of a large amount of unliquefied gaseous refrigerant inside the receiver, causing numerous technical drawbacks and severely restricting the overall performance and operational reliability of the refrigeration unit. Firstly, the large amount of gaseous refrigerant occupies the effective volume inside the receiver, significantly reducing the storage space for liquid refrigerant. This results in a substantial decrease in the effective liquid storage utilization rate of the receiver, failing to fully utilize its refrigerant energy storage and flow regulation functions. Consequently, insufficient refrigerant supply occurs during system load fluctuations, leading to unstable cooling output from the unit. Secondly, the mixing of gas and liquid and the accumulation of gas within the receiver cause an abnormal increase in the high-pressure side pressure of the system. This not only increases the operating load on the compressor, leading to a significant increase in unit energy consumption, but also causes problems such as pipeline pressure fluctuations and large swings in the pressure gauge pointer, exacerbating wear and tear on pipelines and valves.

[0005] Meanwhile, the presence of a large amount of gaseous refrigerant in the receiver will cause uneven refrigerant phases between the gas and liquid phases entering the throttling device. The gaseous refrigerant will mix with the liquid refrigerant and enter the subsequent evaporation heat exchange stage, significantly reducing the heat exchange efficiency of the evaporator. This results in a slower cooling rate and decreased temperature control accuracy, failing to meet the requirements for high-precision and high-stability refrigeration operations. Furthermore, prolonged operation with mixed gas and liquid phases will cause refrigerant circulation disorder in the system, easily leading to compressor return abnormalities, frequent start-stop cycles, and other malfunctions. This will shorten the service life of the compressor and the entire unit, increasing equipment maintenance costs and downtime failure rates.

[0006] In summary, existing refrigeration units suffer from poor condenser adaptability and insufficient liquefaction efficiency, leading to gas accumulation in the liquid receiver and low liquid utilization. This results in a series of problems such as low refrigeration efficiency, high energy consumption, and poor operational stability. The industry urgently needs a refrigeration unit and its supporting control method that can effectively optimize condensation, eliminate gas accumulation in the liquid receiver, and improve the effective utilization rate of the liquid receiver. Summary of the Invention

[0007] This invention addresses the problems existing in the liquid receivers of existing refrigeration units by proposing a refrigeration unit with an ingenious design, simple structure, and better utilization of the liquid receiver, as well as its control method.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The present invention provides a refrigeration unit that can effectively utilize a liquid receiver, including a liquid receiver, a cooler, a compressor, and a condenser. A pressure regulating pipe is provided on the top of the liquid receiver, and a heat exchanger is provided at the end of the pressure regulating pipe away from the liquid receiver. The heat source inlet pipe of the heat exchanger is connected to the pressure regulating pipe, the heat source outlet pipe of the heat exchanger is connected to the liquid inlet of the liquid receiver, the air outlet of the cooler is connected to the heat source inlet of the heat exchanger, the heat source outlet of the heat exchanger is connected to the compressor, and a control expansion valve is provided on the pressure regulating pipe.

[0009] Preferably, a gas-liquid separator is also provided between the heat exchanger and the compressor.

[0010] Preferably, a control method for a refrigeration unit that can effectively utilize a liquid receiver includes the following steps:

[0011] S1. Collect the outlet temperature and outlet pressure of the heat exchanger at a set sampling frequency, and filter the collected raw signals in sequence to obtain the filtered state signals.

[0012] S2. Calculate the heat exchanger heat source outlet superheat based on the filtered state signal: ,in, It is the outlet temperature of the heat exchanger after filtering. This represents the filtered outlet pressure of the heat exchanger. To correspond to pressure The refrigerant saturation temperature is obtained in real time by querying the refrigerant property table, and the superheat value is set.

[0013] S3. Construct an active disturbance rejection controller structure, including three modules: a tracking differentiator, an extended state observer, and a nonlinear error feedback control law. Determine the combination of controller parameters to be optimized, and set the objective function and parameter value range for parameter optimization.

[0014] S4. Improve the dual-subgroup memory cooperative bat algorithm by decomposing the bat population into a global exploration subgroup and a local development subgroup. Construct a trajectory memory bank for each bat in the local development subgroup and update its position using a dual-mode approach of memory recombination and neighborhood search. Set up a knowledge transfer mechanism between subgroups and an adaptive allocation mechanism of computing resources based on the fitness improvement rate to obtain the optimal parameter combination.

[0015] S5. Update the active disturbance rejection controller using the optimal parameter combination, calculate the control quantity of the expansion valve opening degree through the active disturbance rejection controller, and output the control quantity to control the expansion valve after limiting and smoothing the control quantity.

[0016] Preferably, the specific implementation of filtering the acquired raw signals in step S1 to obtain the filtered state signals is as follows:

[0017] S11. Perform zero-mean normalization on the raw heat exchanger outlet temperature signal T(n) and the raw heat exchanger outlet pressure signal P(n) to obtain the normalized temperature signal. and normalized pressure signal Calculate the cross-correlation function of the two normalized signals, and take the time delay corresponding to the peak value of the cross-correlation function as the estimated value of the inherent time delay. The normalized pressure signal is then time-shifted and aligned to obtain the aligned pressure signal. ;

[0018] S12. Calculate the spectral energy density of the normalized temperature signal and the aligned pressure signal, respectively. and Construct the joint spectral energy density function of the frequency-point product of the spectral energy densities of the two signals. Local maxima peaks in the joint spectral energy density function are detected, and the frequency valleys between adjacent peaks are used as frequency band segmentation boundaries to adaptively determine the number of decomposition layers K and the boundaries of each frequency band. Based on the frequency band boundaries, empirical wavelet transforms are performed on the normalized temperature signal and the aligned pressure signal respectively to obtain K layers of temperature mode components. and K-layer pressure modal components ;

[0019] S13. Based on the physical property parameters of the working fluid used in the refrigeration unit, calculate the volume expansion coefficient of any given pair of signals under the current operating conditions. and isothermal compressibility This leads to the thermodynamic coupling coefficient: Define the local thermodynamic consistency residual for any given pair of estimated signals: ;

[0020] S14. Calculate the normalized cross-correlation coefficient between the k-th layer degree modal component and the k-th layer pressure modal component as a coupling consistency index: Calculate the mean of all modal coupling consistency indices. and standard deviation Set dynamic classification threshold and dynamic segmentation threshold: and ,in With sensitivity coefficients ranging from 1.2 to 1.5, all modes are categorized into three types based on the coupling consistency index: For strongly coupled modes, all are retained; Uncoupled modes are all eliminated; the rest are weakly coupled modes and proceed to the next step.

[0021] S15. For each layer of weakly coupled modes, construct a regularized optimization problem that includes signal fidelity terms and thermodynamic consistency constraints, and solve it to obtain the optimized temperature mode components and pressure mode components.

[0022] S16. Superimpose the retained strongly coupled mode, the optimized weakly coupled mode, and the residual component to obtain the reconstructed normalized temperature and pressure signals; perform inverse normalization on the reconstructed signals to obtain the filtered heat exchanger outlet temperature and pressure signals; calculate the global thermodynamic consistency residual of the reconstructed signals. If the global thermodynamic consistency residual is greater than the preset tolerance, update the sensitivity coefficient and return to step S14 to re-execute until the requirements are met.

[0023] Preferably, step S3 constructs an active disturbance rejection controller structure, including three modules: a tracking differentiator, an extended state observer, and a nonlinear error feedback control law. The specific implementation of determining the controller parameter combination to be optimized and setting the objective function and parameter value range for parameter optimization is as follows:

[0024] S31. Construct three core modules for a standard active disturbance rejection controller: a tracking differentiator to smoothly track the target superheat value and extract its differential signal, avoiding setpoint jumps that could impact the system; a third-order extended state observer to estimate the true superheat value, the rate of change of superheat, and the total system disturbance including unmodeled dynamics, parameter perturbations, and external disturbances in real time; and a nonlinear error feedback control law to calculate the basic control quantity based on the tracking error and the differential error.

[0025] S32. Determine the controller parameter vector to be optimized: This includes the tracking speed factor r and filter factor h of the tracking differentiator, and the gain of the extended state observer. and nonlinear function parameters , and the proportional gain of the nonlinear error feedback control law. and differential gain A total of 9 key parameters are combined in module order to form a vector combination of parameters to be optimized;

[0026] S33. Set the comprehensive objective function for parameter optimization, and establish an objective function J with superheat control accuracy, valve action smoothness, and system stability as comprehensive indicators. The expression of the objective function is: Where ITAE is the integral of time multiplied by the absolute error. ,in, This represents the actual superheat. The target overheat level is T, and the total duration is T. To integrate the rate of change of the control quantity, As a system stability indicator, penalties are imposed for overheating and overshoot: Before combining the three elements, standardization is required, and then the elements are combined according to the corresponding preset weight coefficients. The objective function is obtained by performing a weighted summation, and then optimized within the set range of parameter values ​​to minimize the objective function.

[0027] Preferably, step S4 improves the dual-subgroup memory cooperative bat algorithm by decomposing the bat population into a global exploration subgroup and a local development subgroup. A trajectory memory bank is constructed for each bat in the local development subgroup, and its position is updated using a dual-mode approach of memory recombination and neighborhood search. A knowledge transfer mechanism between subgroups and an adaptive allocation mechanism for computational resources based on the fitness improvement rate are set up to obtain the optimal parameter combination. The specific implementation is as follows:

[0028] S41. Initialize the population, setting the size to [size missing]. The bat population was uniformly divided into two functionally complementary subgroups, the first and the second, each containing... / 2 bats; the first subgroup executes a frequency modulation-based global exploration position update strategy, responsible for searching for potential optimal solutions throughout the entire parameter space; the second subgroup executes a trajectory memory reorganization-based local development position update strategy, responsible for fine-grained searching within the neighborhood of high-quality solutions;

[0029] S42. Construct a trajectory memory bank of capacity M for each bat i in the second subgroup, which records the M parameter position vectors with the best fitness in the history of the bat and their corresponding objective function values; the memory bank is dynamically updated using a survival-of-the-fittest mechanism. When a bat obtains a new position and its fitness is better than the worst record in the memory bank, the worst record is replaced; the fitness is the reciprocal of the objective function.

[0030] S43. Update the velocity and position of the bats in the first subgroup according to the following rules: , , ,in, Let be the search frequency of the i-th bat. These are the maximum and minimum frequencies, respectively. A random uniform number in the interval between 0 and 1. Let be the speed of the i-th bat in generation t. Let be the position vector of the i-th bat in generation t. This represents the current globally optimal position vector.

[0031] S44. For bats in the second subgroup, based on the switching probability... Choose either trajectory reconstruction mode or neighborhood fine search mode to update the location;

[0032] S45. Set migration cycle Every Each iteration performs a bidirectional knowledge transfer, injecting the best individual from the first subgroup into the second subgroup to replace the worst-fit individual in the second subgroup; simultaneously, injecting the best individual from the second subgroup into the first subgroup to replace the worst-fit individual in the first subgroup.

[0033] S46. Calculate the fitness improvement rate of each subgroup during the current migration cycle: , ,in, These represent the fitness improvement rates of the first and second subgroups, respectively. Let be the optimal fitness values ​​of the two subgroups in the t-th generation of global exploration. These are the objective function values ​​corresponding to the best individuals at the end of the previous migration cycle for the two subgroups; the number of iterations for each subgroup in the next migration cycle is allocated according to the fitness improvement rate: , ,in, These are the iteration numbers for the first and second subgroups, respectively.

[0034] S47. When the maximum number of iterations is reached or the change in the global optimal fitness value is less than the threshold in two consecutive iterations, the algorithm terminates and outputs the optimal combination of active disturbance rejection controller parameters that minimizes the objective function.

[0035] Preferably, in step S44, the bats in the second population are selected based on the switching probability. The specific implementation of selecting trajectory recombination mode or neighborhood fine search mode to update the position is as follows:

[0036] S441, with probability The trajectory reconstruction mode is used to randomly select two different historical best positions from the bat's trajectory memory database. and New positions are generated through weighted difference recombination: ,in, A uniformly random number within the interval 0 to 1;

[0037] S442, with probability 1- Perform a fine-grained neighborhood search to select the position with the best fitness from the trajectory memory. Perform Lévy flight perturbation within its neighborhood to generate a new location: ,in, All are mutually independent normally distributed random variables. These are the upper and lower bounds of the parameter vector, respectively. This is the adaptive step size scaling factor;

[0038] S443. Select the mode with higher fitness for the new position from the two modes. If the fitness of the new position is better than that of the current position, accept the new position and update the trajectory memory.

[0039] Preferably, the probability The value is adaptively adjusted with the number of iterations, with an initial value of 0.7 and a final value of 0.3. The adjustment formula is as follows: ,in, This represents the maximum number of iterations.

[0040] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0041] 1. This invention provides a refrigeration unit that can effectively utilize a liquid receiver. By installing a pressure regulating pipe at the top of the liquid receiver, the uncondensed high-temperature and high-pressure refrigerant in the liquid receiver enters the heat exchanger, where it serves as a heat source. Meanwhile, the low-temperature and low-pressure gas discharged from the air cooler also enters the air exchanger, where it serves as a heat source. In this way, the uncondensed high-temperature and high-pressure refrigerant, after heat exchange, becomes a high-temperature and high-pressure liquid that enters the liquid receiver, thereby reducing the gas content in the liquid receiver and ensuring the function of the liquid receiver in the refrigeration system. At the same time, the low-temperature and low-pressure gas after heat exchange ensures the absence of liquid, thus eliminating the need for a gas-liquid separator in the refrigeration system and ensuring the performance of the compressor.

[0042] 2. This invention proposes an active disturbance rejection control (ADRC) strategy optimized by an empirical wavelet transform filtering algorithm based on thermodynamic consistency constraints and an improved dual-subgroup memory cooperative bat algorithm. Compared to traditional filtering methods, the filtering algorithm of this invention can not only effectively remove sensor noise and electromagnetic interference, but also ensure that the signal conforms to the physical laws of the refrigerant through thermodynamic consistency constraints. Compared to traditional PID control and conventional ADRC, the control method of this invention automatically tunes the controller parameters through an intelligent optimization algorithm, solving the problems of parameter tuning relying on manual experience and poor adaptability to all operating conditions, and enabling the heat exchanger to always operate in the optimal heat exchange state. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a schematic diagram of the structure of a refrigeration unit that can effectively utilize a liquid receiver, as provided in Example 1.

[0045] Figure 2 This is a schematic flowchart of a control method for a refrigeration unit that can effectively utilize a liquid receiver, as provided in Example 2.

[0046] In the above diagrams, 1 is the liquid receiver; 11 is the pressure regulating pipe; 12 is the expansion control valve; 2 is the air cooler; 3 is the compressor; 4 is the condenser; 5 is the gas-liquid separator; 6 is the oil separator; and 7 is the heat exchanger. Detailed Implementation

[0047] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described below in conjunction with the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0048] Numerous specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways than those described herein, and therefore the invention is not limited to the specific embodiments disclosed in the following specification.

[0049] Example 1, such as Figure 1 As shown, this embodiment aims to provide a refrigeration unit that can effectively utilize a liquid receiver. As is well known, existing refrigeration units mainly include a liquid receiver, a cooler, a compressor, and a condenser, which are connected in a loop. That is, the high-temperature and high-pressure liquid refrigerant in the liquid receiver absorbs heat through the cooler to obtain a low-temperature and low-pressure gaseous refrigerant. The low-temperature and low-pressure gaseous refrigerant is then compressed by the compressor to obtain a high-temperature and high-pressure gaseous refrigerant. The high-temperature and high-pressure gaseous refrigerant is then condensed by the condenser to obtain a high-temperature and high-pressure liquid refrigerant, which then enters the liquid receiver.

[0050] This constitutes the structure of the existing refrigeration unit. Of course, it also requires components such as a gas-liquid separator, oil separator, expansion valve, and throttle valve to complete the refrigeration. In the existing refrigeration unit structure, because the condenser is difficult or impossible to condense, the high-temperature, high-pressure liquid refrigerant entering the receiver also contains a large amount of high-temperature, high-pressure gaseous refrigerant. To ensure the operation of the entire refrigeration unit, the volume of the existing receiver is generally larger than the liquid storage capacity to ensure normal operation. The low-temperature, low-pressure gaseous refrigerant processed by the air cooler also contains a certain amount of low-temperature, low-pressure liquid refrigerant. If the liquid refrigerant enters the compressor, it will damage the compressor valves and pistons. Therefore, a gas-liquid separator is installed. Considering that the gas in the receiver is a high-temperature, high-pressure gaseous refrigerant, and that liquid absorbs heat and turns into gas, and gas releases heat and turns into liquid, a pressure regulating pipe is installed at the top of the receiver in this embodiment. Since the gas is at the top in the receiver, a pressure regulating pipe needs to be installed at the top of the receiver to ensure that only gas is discharged.

[0051] A heat exchanger is installed at the end of the pressure regulating pipe furthest from the liquid receiver. The heat exchanger is designed to facilitate heat transfer between the two fluids, achieving both heating and cooling. The high-temperature, high-pressure gaseous refrigerant in the liquid receiver needs to be cooled, while the low-temperature, low-pressure gaseous refrigerant processed by the air cooler also needs to be heated to convert its low-temperature, low-pressure liquid refrigerant into a low-temperature, low-pressure gaseous refrigerant. Therefore, the heat source inlet pipe of the heat exchanger is connected to the pressure regulating pipe, and the heat source outlet pipe of the heat exchanger is connected to the liquid inlet of the liquid receiver. More specifically, the heat source outlet pipe of the heat exchanger is connected to the pipeline between the condenser and the liquid receiver. This is because the pressure inside the liquid receiver is relatively high, and the heat source outlet pipe of the heat exchanger cannot directly return to the liquid receiver. Therefore, it enters the liquid receiver together with the high-temperature, high-pressure liquid refrigerant discharged from the condenser.

[0052] Instead of directly connecting the evaporative cooler to the compressor (gas-liquid separator), the cooler's outlet is connected to the heat exchanger's heat source inlet, while the heat exchanger's outlet is connected to the compressor. This way, the high-temperature, high-pressure gaseous refrigerant in the receiver releases heat, transforming the low-temperature, low-pressure liquid refrigerant within the low-temperature, low-pressure gaseous refrigerant into a low-temperature, low-pressure gaseous refrigerant. This eliminates the need for a gas-liquid separator, ensuring that only low-temperature, low-pressure gaseous refrigerant enters the compressor. Furthermore, the high-temperature, high-pressure gaseous refrigerant, after releasing heat, becomes a high-temperature, high-pressure liquid refrigerant, increasing the liquid content in the receiver. This avoids the need for large receivers in existing refrigeration units, reducing production costs.

[0053] Considering the different ambient temperatures, it is necessary to control the air intake on the pressure regulating pipe to ensure the heat exchange effect. In this embodiment, a control expansion valve is provided on the pressure regulating pipe.

[0054] To provide redundancy for the entire refrigeration unit and ensure the safe operation of the compressor, a gas-liquid separator is also installed between the heat exchanger and the compressor in this embodiment. The main function of this gas-liquid separator is to prevent low-temperature, low-pressure liquid refrigerant from entering the compressor and causing damage in the event of an accident or insufficient heat exchange.

[0055] Example 2, as Figure 2 As shown, this embodiment provides a specific control method for controlling the expansion valve to ensure optimal heat exchange efficiency. Considering that the optimal control effect of the expansion valve is to ensure that all refrigerant entering the compressor is low-temperature, low-pressure gas, and that as much liquid as possible is discharged from the heat source of the heat exchanger, the control method in this embodiment is to precisely control the opening and closing degree of the expansion valve on the pressure regulating pipe of the receiver. This allows suitable high-temperature, high-pressure gaseous refrigerant to be used to convert the residual low-temperature, low-pressure liquid refrigerant obtained from the air cooler into gaseous refrigerant, minimizing the need for a gas-liquid separator and ensuring that only low-temperature, low-pressure gaseous refrigerant enters the compressor, thus avoiding the possibility of compressor damage.

[0056] Specifically, the heat exchanger outlet temperature and pressure are first collected at a set sampling frequency. The collected raw signals are then filtered sequentially to obtain filtered state signals. Specifically, the collected raw heat exchanger outlet temperature signal T(n) and raw heat exchanger outlet pressure signal P(n) are respectively subjected to zero-mean normalization to obtain normalized temperature signals. and normalized pressure signal Calculate the cross-correlation function of the two normalized signals, and take the time delay corresponding to the peak value of the cross-correlation function as the estimated value of the inherent time delay. The normalized pressure signal is then time-shifted and aligned to obtain the aligned pressure signal. Calculate the spectral energy density of the normalized temperature signal and the aligned pressure signal, respectively. and Construct the joint spectral energy density function of the frequency-point product of the spectral energy densities of the two signals. Local maxima peaks in the joint spectral energy density function are detected, and the frequency valleys between adjacent peaks are used as frequency band segmentation boundaries to adaptively determine the number of decomposition layers K and the boundaries of each frequency band. Based on the frequency band boundaries, empirical wavelet transforms are performed on the normalized temperature signal and the aligned pressure signal respectively to obtain K layers of temperature mode components. and K-layer pressure modal components .

[0057] Next, based on the inherent thermodynamic laws of the refrigerant, a physical constraint correlation between temperature and pressure signals is constructed. According to the standard physical property parameters of the refrigerant used in the refrigeration unit, combined with the current operating conditions, the volume expansion coefficient and isothermal compressibility under the corresponding conditions are calculated in real time, deriving the thermodynamic coupling coefficient and establishing a physical correlation model between temperature and pressure signals. Then, based on this coupling coefficient, a local thermodynamic consistency residual is defined to quantify the degree of deviation between the acquired signals and the basic thermodynamic laws. Specifically, based on the physical property parameters of the refrigerant used in the refrigeration unit, the volume expansion coefficient of any given pair of calculated signals under the current operating conditions is calculated. and isothermal compressibility This leads to the thermodynamic coupling coefficient: Define the local thermodynamic consistency residual for any given pair of estimated signals: This design breaks through the limitations of traditional filtering that relies solely on mathematical algorithms for noise reduction while ignoring physical mechanisms. It allows the signal processing process to closely match the actual thermodynamic characteristics of the refrigeration system, avoiding signal distortion caused by purely mathematical filtering. The thermodynamic consistency residual can accurately distinguish between effective signals and noise interference, significantly improving the reliability and authenticity of the original sensing signal.

[0058] Then, considering the shortcomings of traditional signal processing that relies solely on mathematical features and ignores physical mechanisms, and given the fixed physical relationship between temperature and pressure under refrigeration conditions, the modal components after empirical wavelet decomposition are mixed with effective information and noise, making them indistinguishable solely by mathematical rules, a normalized cross-correlation coefficient is used as the coupling consistency index. Combined with the mean and standard deviation, thresholds are dynamically set to classify modes into three categories: strongly coupled, weakly coupled, and uncoupled, achieving precise classification driven by physical factors. The normalized cross-correlation coefficient between the k-th layer temperature modal component and the k-th layer pressure modal component is calculated as the coupling consistency index. Calculate the mean of all modal coupling consistency indices. and standard deviation Set dynamic classification threshold and dynamic segmentation threshold: and ,in With sensitivity coefficients ranging from 1.2 to 1.5, all modes are categorized into three types based on the coupling consistency index: For strongly coupled modes, all are retained; Uncoupled modes are all eliminated; the remaining weakly coupled modes proceed to the next step. This design approach uses a dynamic threshold that adaptively matches different operating conditions, eliminating the need for repeated manual parameter adjustments. Based on thermodynamic coupling characteristics, the screening method better reflects the actual system mechanism than purely mathematical filtering, accurately eliminating unrelated interference. Hierarchical processing retains high-confidence effective modes and eliminates invalid components, reducing subsequent computational load and improving processing efficiency. Simultaneously, it clarifies the optimization range of weakly coupled modes, laying a solid foundation for signal reconstruction and high-precision superheat calculation, ensuring the stability and reliability of the expansion valve control.

[0059] Furthermore, considering the selected weakly coupled modes, which contain valid operating condition information but do not fully conform to thermodynamic coupling laws, directly eliminating them would result in the loss of key data, while directly retaining them would cause signal distortion. Therefore, a regularized optimization model combining a signal fidelity term and a thermodynamic consistency constraint term is constructed. The fidelity term ensures that the optimized signal does not deviate from the original valid information, while the thermodynamic constraint term forces the signal to conform to the physical laws of the refrigerant. An adaptive step-size gradient descent method is then used to solve the model, ensuring accurate convergence of the optimization results. For each layer of weakly coupled modes, a regularized optimization problem including a signal fidelity term and a thermodynamic consistency constraint term is constructed: It consists of two weighted summations: the first part is a signal fidelity term, which calculates the sum of squares of the differences between the optimized mode and the original mode at corresponding sampling points, ensuring that the optimized mode does not deviate excessively from the original valid signal; the second part is a thermodynamic consistency constraint term, which calculates the sum of squares of the local thermodynamic consistency residuals for each sampling point. The thermodynamic coupling coefficient for each sampling point needs to be dynamically generated based on the temperature and pressure values ​​of the current iteration, by querying the refrigerant's physical property table in real time. An adaptive step-size gradient descent method is used to solve this convex optimization problem. In each iteration, the partial derivatives of the objective function with respect to the temperature and pressure optimization variables are calculated first, and then the mode values ​​are updated along the negative gradient direction. The initial step size is set to 0.01. If the objective function value decreases in this iteration, the step size is multiplied by 1.1; otherwise, it is multiplied by 0.5. When the difference between the objective function values ​​of two consecutive iterations is less than the convergence threshold, or the maximum number of iterations is reached, the iteration terminates and the final optimized temperature and pressure mode components are output. This method can balance signal authenticity and physical rationality, solving the problem of unstable weakly coupled modes; the adaptive gradient descent solution is highly efficient and stable, and can quickly output the optimal modal components; the thermodynamic coupling coefficient is dynamically calculated point by point, which can be adapted to the variable operating conditions of the refrigeration unit; the optimized modes greatly improve the signal reconstruction accuracy, providing reliable data for accurate superheat calculation and high-precision control of the expansion valve.

[0060] The retained strongly coupled modes, the optimized weakly coupled modes, and the residual components are superimposed to obtain the reconstructed normalized temperature and pressure signals. The reconstructed signals are then denormalized to obtain the filtered heat exchanger outlet temperature and pressure signals. The global thermodynamic consistency residual of the reconstructed signals is calculated. If the global thermodynamic consistency residual is greater than the preset tolerance, the sensitivity coefficient is updated and the normalized cross-correlation coefficient calculation step is returned to be re-executed until the requirements are met.

[0061] Next, the superheat at the heat exchanger heat source outlet is calculated based on the filtered state signal, and the target superheat value is set: ,in, It is the outlet temperature of the heat exchanger after filtering. This represents the filtered outlet pressure of the heat exchanger. To correspond to pressure The refrigerant saturation temperature is obtained in real time through the refrigerant property table, and a superheat value is set. This superheat value is a control variable to ensure that a suitable high-temperature, high-pressure gaseous refrigerant is used to convert the residual low-temperature, low-pressure liquid refrigerant obtained from the air cooler into a gaseous refrigerant, minimizing the need for a gas-liquid separator and ensuring that only low-temperature, low-pressure gaseous refrigerant enters the compressor, thus avoiding the possibility of compressor damage.

[0062] An active disturbance rejection controller (ADRC) structure is constructed, comprising three modules: a tracking differentiator, an extended state observer, and a nonlinear error feedback control law. The controller parameter combination to be optimized is determined, and the objective function and parameter range for optimization are set. Specifically, three core modules of a standard ADRC are constructed: a tracking differentiator is built to smoothly track the target superheat value and extract its differential signal, avoiding setpoint jumps that could impact the system; a third-order extended state observer is built to estimate the true superheat value, the rate of change of superheat, and the total system disturbance including unmodeled dynamics, parameter perturbations, and external disturbances in real time; and a nonlinear error feedback control law is built to calculate the basic control variables based on the tracking error and the differential error. The controller parameter vector to be optimized is determined by combining the tracking speed factor r and the filter factor h of the tracking differentiator, and the gain of the extended state observer. and nonlinear function parameters , and the proportional gain of the nonlinear error feedback control law. and differential gain Nine key parameters are combined sequentially by module to form a parameter vector combination to be optimized. Among them, the tracking differentiator (TD) module has two parameters: the tracking speed factor r determines the tracking speed of the tracking differentiator to the target value; the larger r is, the faster the tracking, but excessively large r can easily cause overshoot and noise amplification. The filtering factor h determines the filtering capability of the tracking differentiator; the larger h is, the better the filtering effect, but it will increase tracking lag.

[0063] The third-order extended state observer (ESO) module has five parameters. The extended state observer is the core of ADRC, used to estimate the true superheat value, the rate of change of superheat, and the total system disturbance (including unmodeled dynamics, parameter perturbations, and external disturbances) in real time. Observer gain. Physical meaning: The correction gain for superheat state estimation affects the convergence speed of the state estimation. Observer gain. Physical meaning: The correction gain for estimating the rate of change of superheat affects the estimation accuracy of the differential signal. Observer gain. Physical meaning: The correction gain for the total system disturbance estimation, affecting the speed and accuracy of the disturbance estimation. Nonlinear function parameters. Physical meaning: The exponent of the first-order nonlinear power function of ESO, usually taken as 0.5, is used to improve the estimation sensitivity under small errors. Nonlinear function parameters. Physical meaning: The linear segment width of the first-order nonlinear function of ESO, used to avoid high-frequency flutter. Nonlinear Error Feedback Control Law (NLSEF) module parameters (2): The nonlinear error feedback control law calculates the basic control quantity based on the tracking error and differential error to achieve closed-loop regulation of superheat. Proportional gain. Physical meaning: The proportional gain for superheat tracking error determines the strength of the control effect. Differential gain. Physical meaning: The differential adjustment gain of the superheat change rate error is used to suppress system overshoot and oscillation.

[0064] A comprehensive objective function for parameter optimization is defined, and an objective function J is established with superheat control accuracy, valve action smoothness, and system stability as comprehensive indicators. The expression of the objective function is as follows: Where ITAE is the integral of time multiplied by the absolute error. ,in, This represents the actual superheat. The target overheat level is T, and the total duration is T. To integrate the rate of change of the control quantity, As a system stability indicator, penalties are imposed for overheating and overshoot: Before combining the three elements, standardization is required, and then the elements are combined according to the corresponding preset weight coefficients. The objective function is obtained by performing a weighted summation, and then optimized within the set range of parameter values ​​to minimize the objective function.

[0065] This design addresses the pain points of difficult tuning of nine key parameters in active disturbance rejection controllers (ADRCs) and premature convergence of conventional optimization algorithms. ADRCs have numerous parameters and strong coupling; traditional manual tuning relies on experience and has poor adaptability to all operating conditions. Conventional bat algorithms suffer from an imbalance between global exploration and local exploitation, are prone to getting trapped in local optima, and have low optimization accuracy, failing to meet the high-precision parameter optimization requirements of refrigeration systems under varying operating conditions. Therefore, this step decomposes the bat algorithm into a dual-subgroup cooperative architecture, introducing a trajectory memory library, dual-mode position updates, subgroup knowledge transfer, and adaptive computational resource allocation mechanisms. With global exploration of the topology space and local development for fine-tuning parameters as the core logic, intelligent optimal tuning of controller parameters is achieved. An improved dual-subgroup memory cooperative bat algorithm is adopted, decomposing the bat population into a global exploration subgroup and a local exploitation subgroup. A trajectory memory library is built for each bat in the local exploitation subgroup, and a dual-mode update of position using memory recombination and neighborhood search is employed. A knowledge transfer mechanism between subgroups and an adaptive computational resource allocation mechanism based on fitness improvement rate are set up to obtain the optimal parameter combination.

[0066] Specifically, first initialize the population, setting its size to [size missing]. The bat population is uniformly divided into two complementary subgroups, the first and second, to resolve the inherent contradiction between global search and local exploration in a single population. Conventional bat algorithms use only a single population for search, making it difficult to simultaneously cover a large parameter space traversal and fine-grained discovery of high-quality solution regions, easily leading to insufficient early-stage search and getting trapped in local optima later. However, the active disturbance rejection controller (ADRC) has a broad optimization space for its nine parameters, requiring first globally locating high-quality solution intervals and then fine-grained local optimization. Therefore, dividing the population into two allows the first subgroup to focus on global exploration and the second subgroup on local exploration, forming a complementary collaborative search architecture. Each subgroup contains... Two bats are used in a cluster; the first subgroup executes a frequency-modulated global exploration position update strategy, responsible for searching for potential optimal solutions across the entire parameter space; the second subgroup executes a trajectory memory reorganization-based local development position update strategy, responsible for fine-tuning within the neighborhood of high-quality solutions. The two subgroups have clear divisions of labor and do not interfere with each other. The global exploration subgroup can quickly cover the entire parameter space, avoiding missing potential optimal solutions; the local development subgroup focuses on fine-tuning high-quality areas, improving parameter tuning accuracy. Uniform clustering ensures reasonable allocation of computational resources with no redundant consumption.

[0067] To address the issues of lost historical high-quality solutions and repeated invalid searches in local search, the core objective of the local development subgroup is to meticulously mine optimal parameters. Relying solely on the current position for searching can easily lead to forgetting high-quality solutions from previous iterations, resulting in chaotic search paths and slow convergence. Optimizing cooling system parameters requires preserving highly adaptable parameter combinations. Therefore, a trajectory memory bank with a fixed capacity is established for each bat, storing historically optimal parameter positions and objective function values ​​according to a survival-of-the-fittest mechanism, ensuring that local searches are always guided by valuable experience. Thus, a trajectory memory bank of capacity M is constructed for each bat i in the second subgroup to record the M parameter position vectors with the best fitness in the bat's history and their corresponding objective function values. The memory bank is dynamically updated using a survival-of-the-fittest mechanism; when a bat obtains a new position and its fitness is better than the worst record in the memory, the worst record is replaced. The fitness is the reciprocal of the objective function. This trajectory memory bank design can effectively preserve high-quality solution information, avoiding repeated searches of invalid regions and significantly improving the efficiency of local development. The dynamic update mechanism ensures that the memory bank always stores the current optimal solution, continuously providing high-quality references for position updates. The deep integration of memory and local search makes the algorithm more directional when performing fine optimization, effectively avoiding local optimum traps and improving the stability and accuracy of parameter optimization.

[0068] Next, considering the need to ensure the global subgroup has a wide-ranging and highly traversable search capability, the core task of the global exploration subgroup is to quickly locate the potential optimal region in the parameter space. Conventional fixed-step search is prone to blind spots and cannot fully cover the combination space of the nine controller parameters. Frequency modulation allows the bat search frequency to be dynamically adjusted within a preset range. Combined with the velocity-position update formula, it drives the subgroup to traverse the entire parameter space without blind spots, quickly selecting the distribution area of ​​high-quality solutions. Specifically, the velocity and position of the bats in the first subgroup are updated according to the following rules: , , ,in, Let be the search frequency of the i-th bat. These are the maximum and minimum frequencies, respectively. A random uniform number in the interval between 0 and 1. Let be the speed of the i-th bat in generation t. Let be the position vector of the i-th bat in generation t. This represents the current globally optimal position vector. This design, guided by the globally optimal position, makes the search direction more targeted and improves global exploration efficiency. Through efficient global exploration, it provides precise search target areas for local development subgroups, significantly improving the overall algorithm's optimization speed.

[0069] Considering that the purpose of the second subgroup is to balance utilizing historical experience from local searches with the ability to escape local optima, relying solely on trajectory reconstruction for local development subgroups can easily limit them to the historical solution space and prevent them from breaking through; while relying solely on neighborhood search can easily lead to blind perturbations and the loss of high-quality information. Cooling system parameter optimization requires fine-tuning while preserving high-quality solutions. Therefore, an adaptive switching probability selection mode is used: trajectory reconstruction utilizes historical optimal solutions from the memory bank to generate new positions, while neighborhood search broadens the search range through Lévy flight perturbations. The optimal mode is chosen from these two. Specifically, for bats in the second subgroup, the switching probability is used... Choose either trajectory reconstruction mode or neighborhood fine search mode to update the position, and further, use probability... The trajectory reconstruction mode is used to randomly select two different historical best positions from the bat's trajectory memory database. and New positions are generated through weighted difference recombination: ,in, A uniformly random number within the interval 0 to 1; with a probability of 1- Perform a fine-grained neighborhood search to select the position with the best fitness from the trajectory memory. Perform Lévy flight perturbation within its neighborhood to generate a new location: ,in, All are mutually independent normally distributed random variables. These are the upper and lower bounds of the parameter vector, respectively. The adaptive step size scaling factor is used; the mode with higher fitness for the new position is selected from the two modes. If the fitness of the new position is better than that of the current position, the new position is accepted and the trajectory memory is updated. Further, the probability mentioned... The value is adaptively adjusted with the number of iterations, with an initial value of 0.7 and a final value of 0.3. The adjustment formula is as follows: ,in, This represents the maximum number of iterations. The dual-mode collaboration fully leverages historical best practices to ensure accurate search direction while avoiding local optima through random perturbations, balancing refinement and innovation. The switching probability adaptively adjusts with each iteration, prioritizing rapid convergence through recombination in the early stages and overcoming bottlenecks through perturbations in the later stages, adapting to the full lifecycle optimization requirements. Lévy's adaptive flight perturbation step size precisely matches the parameter optimization accuracy requirements, resulting in more refined local optimization. This step significantly improves parameter tuning accuracy, ensuring that controller parameters adapt to the dynamic changes of the refrigeration unit under all operating conditions.

[0070] In the following steps, a bidirectional knowledge transfer mechanism between the two subgroups is established. Traditional designs suffer from information silos and overall population performance degradation due to independent evolution of subgroups. If the global and local subgroups search independently for a long period, problems arise such as the global subgroup failing to propagate its superior solution after finding one, and the local subgroup getting stuck in local optima, leading to slow convergence and poor optimization. Therefore, parameter optimization requires co-evolution of the two subgroups. Thus, bidirectional transfer is performed at fixed intervals, injecting the best individual from each subgroup into the other's population and replacing the worst individual, thereby achieving information sharing on superior solutions.

[0071] Specifically, setting a migration cycle Every Each iteration performs a bidirectional knowledge transfer, injecting the best individual from the first subgroup into the second subgroup, replacing the worst-fit individual in the second subgroup; simultaneously, injecting the best individual from the second subgroup into the first subgroup, replacing the worst-fit individual in the first subgroup. Calculate the fitness improvement rate of each subgroup within the current transfer cycle: , ,in, These represent the fitness improvement rates of the first and second subgroups, respectively. Let be the optimal fitness values ​​of the two subgroups in the t-th generation of global exploration. These are the objective function values ​​corresponding to the best individuals at the end of the previous migration cycle for the two subgroups; the number of iterations for each subgroup in the next migration cycle is allocated according to the fitness improvement rate: , ,in, These represent the iteration counts for the first and second subgroups, respectively. The algorithm terminates when the maximum number of iterations is reached or the change in the global optimal fitness value between two consecutive iterations is less than a threshold. The algorithm outputs the optimal combination of active disturbance rejection controller parameters that minimizes the objective function. Here, bidirectional knowledge transfer breaks down information barriers between subgroups, allowing high-quality solutions from global exploration to quickly empower local development, while high-quality solutions from local fine-tuning conversely improve global search efficiency, forming a closed loop of global boundary expansion and collaborative evolution. A fixed migration cycle ensures a moderate information transmission frequency, avoiding both frequent migrations disrupting the search rhythm and long periods without migration leading to information lag. This mechanism significantly improves the overall optimization ability of the population, accelerates algorithm convergence, avoids premature convergence, and makes controller parameter optimization more efficient and accurate.

[0072] Finally, the optimal parameter combination is used to update the active disturbance rejection controller (ADRC). The ADRC calculates the expansion valve opening control quantity, limits and smooths the control quantity, and then outputs it to the expansion valve. The limiting ensures that the control quantity is within a preset range. The flow rate of the high-temperature, high-pressure gaseous refrigerant entering the heat exchanger is determined by controlling the valve opening of the expansion valve. Inside the heat exchanger, the high-temperature, high-pressure gaseous refrigerant acts as a heat source, releasing heat and condensing, while the low-temperature, low-pressure gas-liquid mixture at the outlet of the air cooler acts as a heat source, absorbing heat. By adjusting the heat release through valve opening, the residual liquid refrigerant can be precisely and completely vaporized, eliminating the need for a gas-liquid separator and allowing it to directly enter the compressor, thus minimizing the risk of liquid refrigerant slugging and damaging the compressor.

[0073] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments for application in other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A refrigeration unit capable of efficiently utilizing a liquid accumulator, comprising a liquid accumulator, a cold air generator, a compressor, and a condenser, characterized in that, A pressure regulating pipe is provided at the top of the liquid storage tank. A heat exchanger is provided at the end of the pressure regulating pipe away from the liquid storage tank. The heat source inlet pipe of the heat exchanger is connected to the pressure regulating pipe. The heat source outlet pipe of the heat exchanger is connected to the liquid inlet of the liquid storage tank. The air outlet of the air cooler is connected to the heat source inlet of the heat exchanger. The heat source outlet of the heat exchanger is connected to the compressor. A control expansion valve is provided on the pressure regulating pipe.

2. The refrigeration unit capable of effectively utilizing a liquid accumulator according to claim 1, characterized in that, A gas-liquid separator is also provided between the heat exchanger and the compressor.

3. The control method for a refrigeration unit that can effectively utilize a liquid receiver as described in claim 1, characterized in that, Includes the following steps: S1. Collect the outlet temperature and outlet pressure of the heat exchanger at a set sampling frequency, and filter the collected raw signals in sequence to obtain the filtered state signals. S2. Calculate the heat exchanger heat source outlet superheat based on the filtered state signal: ,in, It is the outlet temperature of the heat exchanger after filtering. This represents the filtered outlet pressure of the heat exchanger. To correspond to pressure The refrigerant saturation temperature is obtained in real time by querying the refrigerant property table, and the superheat value is set. S3. Construct an active disturbance rejection controller structure, including three modules: a tracking differentiator, an extended state observer, and a nonlinear error feedback control law. Determine the combination of controller parameters to be optimized, and set the objective function and parameter value range for parameter optimization. S4. Improve the dual-subgroup memory cooperative bat algorithm by decomposing the bat population into a global exploration subgroup and a local development subgroup. Construct a trajectory memory bank for each bat in the local development subgroup and update its position using a dual-mode approach of memory recombination and neighborhood search. Set up a knowledge transfer mechanism between subgroups and an adaptive allocation mechanism of computing resources based on the fitness improvement rate to obtain the optimal parameter combination. S5. Update the active disturbance rejection controller using the optimal parameter combination, calculate the control quantity of the expansion valve opening degree through the active disturbance rejection controller, and output the control quantity to control the expansion valve after limiting and smoothing the control quantity.

4. The control method for a refrigeration unit that can effectively utilize a liquid receiver according to claim 3, characterized in that, The specific implementation of filtering the acquired raw signals in step S1 to obtain the filtered state signals is as follows: S11. Perform zero-mean normalization on the raw heat exchanger outlet temperature signal T(n) and the raw heat exchanger outlet pressure signal P(n) to obtain the normalized temperature signal. and normalized pressure signal Calculate the cross-correlation function of the two normalized signals, and take the time delay corresponding to the peak value of the cross-correlation function as the estimated value of the inherent time delay. The normalized pressure signal is then time-shifted and aligned to obtain the aligned pressure signal. ; S12. Calculate the spectral energy density of the normalized temperature signal and the aligned pressure signal, respectively. and Construct the joint spectral energy density function of the frequency-point product of the spectral energy densities of the two signals. Local maxima peaks in the joint spectral energy density function are detected, and the frequency valleys between adjacent peaks are used as frequency band segmentation boundaries to adaptively determine the number of decomposition layers K and the boundaries of each frequency band. Based on the frequency band boundaries, empirical wavelet transforms are performed on the normalized temperature signal and the aligned pressure signal respectively to obtain K layers of temperature mode components. and K-layer pressure modal components ; S13. Based on the physical property parameters of the working fluid used in the refrigeration unit, calculate the volume expansion coefficient of any given pair of signals under the current operating conditions. and isothermal compressibility This leads to the thermodynamic coupling coefficient: Define the local thermodynamic consistency residual for any given pair of estimated signals: ; S14. Calculate the normalized cross-correlation coefficient between the k-th layer degree modal component and the k-th layer pressure modal component as a coupling consistency index: Calculate the mean of all modal coupling consistency indices. and standard deviation Set dynamic classification threshold and dynamic segmentation threshold: and ,in With sensitivity coefficients ranging from 1.2 to 1.5, all modes are categorized into three types based on the coupling consistency index: For strongly coupled modes, all are retained; Uncoupled modes are all eliminated; The remaining modes are weakly coupled and will proceed to the next step of processing. S15. For each layer of weakly coupled modes, construct a regularized optimization problem that includes signal fidelity terms and thermodynamic consistency constraints, and solve it to obtain the optimized temperature mode components and pressure mode components. S16. Superimpose the retained strongly coupled mode, the optimized weakly coupled mode, and the residual component to obtain the reconstructed normalized temperature and pressure signals; perform inverse normalization on the reconstructed signals to obtain the filtered heat exchanger outlet temperature and pressure signals; calculate the global thermodynamic consistency residual of the reconstructed signals. If the global thermodynamic consistency residual is greater than the preset tolerance, update the sensitivity coefficient and return to step S14 to re-execute until the requirements are met.

5. The control method for a refrigeration unit that can effectively utilize a liquid receiver according to claim 3, characterized in that, Step S3 involves constructing an active disturbance rejection controller structure, comprising three modules: a tracking differentiator, an extended state observer, and a nonlinear error feedback control law. The specific implementation of determining the controller parameter combination to be optimized and setting the objective function and parameter value range for parameter optimization is as follows: S31. Construct three core modules for a standard active disturbance rejection controller: a tracking differentiator to smoothly track the target superheat value and extract its differential signal, avoiding setpoint jumps that could impact the system; a third-order extended state observer to estimate the true superheat value, the rate of change of superheat, and the total system disturbance including unmodeled dynamics, parameter perturbations, and external disturbances in real time; and a nonlinear error feedback control law to calculate the basic control quantity based on the tracking error and the differential error. S32. Determine the controller parameter vector to be optimized: This includes the tracking speed factor r and filter factor h of the tracking differentiator, and the gain of the extended state observer. and nonlinear function parameters , and the proportional gain of the nonlinear error feedback control law. and differential gain A total of 9 key parameters are combined in module order to form a vector combination of parameters to be optimized; S33. Set the comprehensive objective function for parameter optimization, and establish an objective function J with superheat control accuracy, valve action smoothness, and system stability as comprehensive indicators. The expression of the objective function is: Where ITAE is the integral of time multiplied by the absolute error. ,in, This represents the actual superheat. The target overheat level is T, and the total duration is T. To integrate the rate of change of the control quantity, As a system stability indicator, penalties are imposed for overheating and overshoot: Before combining the three elements, standardization is required, and then the elements are combined according to the corresponding preset weight coefficients. The objective function is obtained by performing a weighted summation, and then optimized within the set range of parameter values ​​to minimize the objective function.

6. The control method for a refrigeration unit that can effectively utilize a liquid receiver according to claim 3, characterized in that, The improved dual-subgroup memory cooperative bat algorithm in step S4 decomposes the bat population into a global exploration subgroup and a local development subgroup. A trajectory memory bank is constructed for each bat in the local development subgroup, and its position is updated using a dual-mode approach of memory recombination and neighborhood search. A knowledge transfer mechanism between subgroups and an adaptive allocation mechanism for computational resources based on the fitness improvement rate are set up to obtain the optimal parameter combination. The specific implementation is as follows: S41. Initialize the population, setting the size to [size missing]. The bat population was uniformly divided into two functionally complementary subgroups, the first and the second, each containing... / 2 bats; The first subgroup executes a frequency modulation-based global exploration position update strategy, responsible for searching for potential optimal solutions throughout the entire parameter space; the second subgroup executes a trajectory memory reorganization-based local development position update strategy, responsible for fine-grained searching within the neighborhood of high-quality solutions. S42. Construct a trajectory memory bank of capacity M for each bat i in the second subgroup, which records the M parameter position vectors with the best fitness in the history of the bat and their corresponding objective function values; the memory bank is dynamically updated using a survival-of-the-fittest mechanism. When a bat obtains a new position and its fitness is better than the worst record in the memory bank, the worst record is replaced; the fitness is the reciprocal of the objective function. S43. Update the velocity and position of the bats in the first subgroup according to the following rules: , , ,in, Let be the search frequency of the i-th bat. These are the maximum and minimum frequencies, respectively. A random uniform number in the interval between 0 and 1. Let be the speed of the i-th bat in generation t. Let be the position vector of the i-th bat in generation t. This represents the current globally optimal position vector. S44. For bats in the second subgroup, based on the switching probability... Choose either trajectory reconstruction mode or neighborhood fine search mode to update the location; S45. Set migration cycle Every Each iteration performs a bidirectional knowledge transfer, injecting the best individual from the first subgroup into the second subgroup to replace the worst-fit individual in the second subgroup; simultaneously, injecting the best individual from the second subgroup into the first subgroup to replace the worst-fit individual in the first subgroup. S46. Calculate the fitness improvement rate of each subgroup during the current migration cycle: , ,in, These represent the fitness improvement rates of the first and second subgroups, respectively. Let be the optimal fitness values ​​of the two subgroups in the t-th generation of global exploration. These are the objective function values ​​corresponding to the best individuals at the end of the previous migration cycle for the two subgroups; the number of iterations for each subgroup in the next migration cycle is allocated according to the fitness improvement rate: , ,in, These are the iteration numbers for the first subgroup and the second subgroup, respectively. S47. When the maximum number of iterations is reached or the change in the global optimal fitness value is less than the threshold in two consecutive iterations, the algorithm terminates and outputs the optimal combination of active disturbance rejection controller parameters that minimizes the objective function.

7. The control method for a refrigeration unit that can effectively utilize a liquid receiver according to claim 6, characterized in that, In step S44, the bats in the second population are processed according to the switching probability. The specific implementation of selecting trajectory recombination mode or neighborhood fine search mode to update the position is as follows: S441, with probability The trajectory reconstruction mode is used to randomly select two different historical best positions from the bat's trajectory memory database. and New positions are generated through weighted difference recombination: ,in, A uniformly random number within the interval 0 to 1; S442, with probability 1- Perform a fine-grained neighborhood search to select the position with the best fitness from the trajectory memory. Perform Lévy flight perturbation within its neighborhood to generate a new location: ,in, All are mutually independent normally distributed random variables. These are the upper and lower bounds of the parameter vector, respectively. This is the adaptive step size scaling factor; S443. Select the mode with higher fitness for the new position from the two modes. If the fitness of the new position is better than that of the current position, accept the new position and update the trajectory memory.

8. A control method for a refrigeration unit that can effectively utilize a liquid receiver according to claim 7, characterized in that, The probability The value is adaptively adjusted with the number of iterations, with an initial value of 0.7 and a final value of 0.

3. The adjustment formula is as follows: ,in, This represents the maximum number of iterations.

Images