A self-adaptive intelligent control method for a cascade high-temperature heat pump

Abstract

The application discloses a kind of self-adapting intelligent control method of cascade high-temperature heat pump, by constructing the liquid film dynamic observer of falling-film evaporator liquid film containing liquid film breakage loss term, realize the real-time soft measurement of liquid film thickness and dry spot coverage rate;By constructing system dynamic prediction model, based on liquid film thickness, dry spot coverage rate and the operating state of cascade high-temperature heat pump, predict the core operating state of cascade high-temperature heat pump in future setting time domain;By establishing the heat transfer model of plate shell heat exchanger microelement along the way, the pinch risk is quantified;Collaborative control is carried out using multi-objective rolling optimization with soft constraint quadratic penalty, and the multi-mode adaptive switching threshold is derived based on system thermal inertia.COP is increased by 8~12% compared with conventional PID control, dry spot coverage rate is stably controlled within 5%, the minimum temperature difference of plate shell heat exchanger is maintained above 1.5K, the outlet water temperature fluctuation is ≤±0.4℃ in all working conditions (cooling water 20-40℃), and no parameter oscillation phenomenon occurs.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology for industrial high-temperature heat pump systems, and in particular relates to an adaptive intelligent control method for cascade high-temperature heat pumps. Background Technology

[0002] In industrial production processes such as beer, beverage, and food processing, a large amount of low-temperature cooling water (20-40°C) is generated, while a large amount of high-temperature hot water (above 90°C) is needed for sterilization, cleaning, and cooking. Using a cascade high-temperature heat pump to recover low-temperature waste heat and produce high-temperature hot water is an effective technical means. Specifically, a low-temperature stage and a high-temperature stage are coupled through a condenser-evaporator. The low-temperature stage recovers low-grade waste heat, which is then transferred to the high-temperature stage via the condenser-evaporator. After thermal boosting by the high-temperature stage, a high-temperature heat source of 85-130°C is output. Existing cascade high-temperature heat pumps generally use PID control or an integrated black-box controller for overall system regulation. The low-temperature stage evaporator uses a falling film evaporator, and the condenser-evaporator uses a plate-and-shell evaporator-condenser. The refrigerant used is a low-GWP refrigerant, typically represented by R1234ze(E) and R1233zd(E). The working fluid used has characteristics such as low latent heat of vaporization and low evaporation pressure, which leads to the following technical problems in practical applications: the liquid film of the working fluid with low latent heat of vaporization is prone to rupture and form dry spots in falling film evaporators, resulting in a sharp drop in the heat transfer coefficient; the working fluid with low evaporation pressure is sensitive to pressure drop along the process in plate and shell evaporators and is prone to local pinch-point risk (the minimum local temperature difference between hot and cold fluids along the length of the plate bundle inside the plate and shell evaporator and the temperature difference is <1K), which makes the heat transfer driving force approach zero.

[0003] Therefore, existing cascade high-temperature heat pump systems have the following technical defects: 1. PID control uses closed-loop control of the electronic expansion valve opening and compressor frequency by detecting the superheat of the low-temperature stage intake air and the outlet water temperature of the high-temperature stage. It can only make lag corrections based on the parameter deviation at the current moment, and cannot predict the liquid film state and pinch-point risks. To avoid liquid film dry spots, the superheat must be conservatively set (≥3K), resulting in a system COP that is more than 10% lower than the theoretical optimal value. Furthermore, PID control cannot handle multi-objective coupled optimization problems, and is prone to parameter oscillations when operating conditions fluctuate.

[0004] 2. Integrated black-box controllers are mostly based on pure data-driven modeling, without incorporating liquid film state and pinch risk into the prediction logic. The model lacks physical mechanism support, resulting in low control accuracy. If the operating conditions fluctuate, the control accuracy will further decrease, and problems such as control oscillation and frequent actuator movements are very likely to occur.

[0005] To overcome the technical defects of the aforementioned PID and integrated black box controllers, some improvement solutions propose using optical and ultrasonic sensors to directly detect the liquid film thickness. However, such solutions are costly, susceptible to interference from oil and wet steam in the heat exchange chamber, and have poor field reliability, making them completely unsuitable for the long-term operation requirements of industrial-grade cascade heat pumps. Summary of the Invention

[0006] The present invention aims to solve the aforementioned technical problems existing in the prior art by providing an adaptive intelligent control method for cascade high-temperature heat pumps.

[0007] The technical solution of this invention is: an adaptive intelligent control method for a cascade high-temperature heat pump, comprising an adaptive intelligent controller, a PID controller, and a cascade high-temperature heat pump consisting of a low-temperature stage and a high-temperature stage. The PID controller has an inner and outer loop cascade structure. A liquid film dynamic observer for a falling film evaporator is established to achieve real-time soft measurement of liquid film thickness and dry spot coverage. A system dynamic prediction model is constructed, and based on real-time soft measurement data and the operating parameters of the cascade high-temperature heat pump, the core operating state of the cascade high-temperature heat pump within a future set time domain is predicted. A micro-element friction heat transfer model of a plate-shell heat exchanger is established to quantify pinch risk during operation. The adaptive intelligent controller receives the core operating state output by the system dynamic prediction model and the pinch risk data quantified by the micro-element friction heat transfer model of the plate-shell heat exchanger. Based on multi-objective rolling optimization with soft constraint secondary penalty, a dynamic setpoint is generated and sent to the outer loop of the PID controller. The outer loop of the PID controller outputs a control quantity according to the dynamic setpoint to regulate the frequency of the low-temperature stage compressor, the frequency of the high-temperature stage compressor, and the opening of the high-temperature stage electronic expansion valve. The inner loop of the PID controller tracks the dynamic setpoint of the outer loop of the PID controller in real time and controls the opening of the electronic expansion valve of the low-temperature stage.

[0008] The preferred cascade high-temperature heat pump includes a plate-and-shell evaporator-condenser connecting a low-temperature stage and a high-temperature stage, wherein the low-temperature stage uses a falling film evaporator, and the control method is implemented according to the following specific steps: S1. The liquid film dynamic observer of the falling film evaporator first collects the operating parameters of the low-temperature stage, and solves the liquid film state prediction value through dynamic equations; then, it solves the superheat deviation and liquid film rupture loss based on the liquid film state prediction value; then, it corrects the liquid film state prediction value based on the superheat deviation and liquid film rupture loss; and iterates to obtain the real-time estimation result; the superheat deviation is the difference between the measured suction superheat and the predicted suction superheat. S2. The system dynamic prediction model takes the current core operating state and liquid film state of the cascade high-temperature heat pump as internal state variables, receives the control commands output by the adaptive intelligent controller as input vectors, and calculates and outputs the core operating state parameters of the cascade high-temperature heat pump in the future set time domain. S3. The friction heat transfer model of the plate-shell evaporator-condenser divides the plate bundle into M micro-elements along the working fluid flow direction. Based on the output of the system dynamic prediction model, the local heat transfer temperature difference between the condensing side and the evaporating side in each micro-element is solved. Then, the ratio of the maximum to the minimum local temperature difference along the plate bundle is calculated. Combined with the exponential penalty term of the minimum local temperature difference, the heat transfer deterioration risk index is obtained. S4. The adaptive intelligent controller is constructed based on the output values ​​of the system dynamic prediction model and the friction heat transfer model of the plate-and-shell evaporator-condenser. It includes a system performance coefficient COP maximization term, a heat transfer deterioration risk index penalty term, a dry spot coverage penalty term, and a regularization term. The heat transfer deterioration risk index penalty term and the dry spot coverage penalty term both adopt a soft-constraint quadratic penalty form, and the penalty is activated only when the risk index or dry spot coverage exceeds its respective threshold.

[0009] The preferred liquid film dynamics observer for the falling film evaporator includes a state prediction equation for the average liquid film thickness and a state prediction equation for the dry spot coverage. The state prediction equation for the average thickness of the liquid film is as follows: ; The state prediction equation for the dry spot coverage is as follows: when hour, ; when d ≥ crit hour, ; In the formula: dδ / dt This represents the rate of change of the average thickness of the liquid film. dφ / dt The rate of change of the dry spot coverage of the liquid film; d This is an estimated average thickness of the liquid film. f This is an estimated value for the dry spot coverage of the liquid film; The refrigerant spray mass flow rate at the outlet of the low-temperature electronic expansion valve; The mass flow rate of refrigerant evaporation within the evaporator is... , For heat exchange in the evaporator, r The latent heat of vaporization of the refrigerant; The loss of mass flow rate is due to liquid film rupture; ; In the formula K break This is the liquid film rupture loss coefficient; ; d crit The critical liquid film thickness is determined by taking the initial spray liquid film thickness. d 0 20-30%; The density of the refrigerant; This refers to the total circumferential wetted area of ​​the evaporator tube bank. ; in W peri For the wetting perimeter of a single tube, N tube For the total number of people under management, L tube For effective management; , For observer gain, Take the negative value. Take positive values; This is the measured value of the suction superheat of the cryogenic compressor; The compressor suction superheat is predicted based on the liquid film state. ; In the formula: K sh The superheat conversion coefficient; Ueff(d,f) · A evap Effective heat transfer conductivity modulated by the liquid film; D T sh ,0 This is a correction term for the baseline superheat. ; In the formula U wet (d) The heat transfer coefficient of the wetted region. U wet (d) = C·(n / n 0 ) (-1 / 3) , C As the reference heat transfer coefficient, d 0 The initial liquid film thickness, ,in This refers to the refrigerant spray flow rate under rated operating conditions. U dry The heat transfer coefficient of the dry spot region is taken as a value. U wet (d)10-30%; d crit The critical liquid film thickness is defined as a value of [value to be filled in]. d 0 20-30%; α dry This is the dry spot spread rate coefficient; -α recover This represents the dry spot recovery rate coefficient.

[0010] The preferred system dynamic prediction model is as follows: dx / dt = A · x + B · u ; In the formula: x For state variables, x = [ P evap , P cond , T mid , d , f ] T The P evap The evaporation pressure at the low temperature stage. P cond For high-temperature condensation pressure, T mid The intermediate temperature; u For the input vector, u = [ f low , f high , v high ] T The f low For the frequency of the cryogenic compressor, f high For high-temperature compressor frequency, v high This refers to the opening degree of the high-temperature electronic expansion valve. A is a 5×5 diagonally dominant matrix, where the diagonal elements reflect the thermal inertia of each component, with a typical value of a. 11 =-1 / t evap a 22 = -1 / t cond a 33 = -1 / tmid The t evap The thermal inertia time constant of the low-temperature stage evaporator, the t cond The thermal inertia time constant of the high-temperature stage condenser, the t mid The time constant of thermal inertia for intermediate temperature stages; off-diagonal elements reflect... T mid and P evap and P cond The coupling relationship; B is the input matrix, which reflects the adjustment gain of the input vector on the state variables.

[0011] Preferably, the heat transfer deterioration risk index is set as follows: Rk, The calculation formula is as follows: Rk = max( ΔTi,k ) / max[min( ΔTi,k ), e ] + λ·exp[-min (ΔTi,k) / ΔT ref ]; In the formula: k This indicates the sequence number of the computation step within the specified time domain for predicting the future. ΔT i For the first i Local heat transfer temperature difference within a micro-element; ; In the formula: T sat This is a saturation temperature function, representing the saturation temperature of the refrigerant at a given pressure; P cond , in This refers to the condensing pressure at the inlet of the low-temperature stage condenser. P evap , in This refers to the evaporation pressure at the inlet of the high-temperature evaporator. SDPcond , i To the condenser side inlet to the first i The sum of the cumulative pressure drops of all infinitesimal elements between each infinitesimal entry point; Σ ΔP evap , i To the evaporator side inlet to the first i The sum of the cumulative pressure drops of all infinitesimal elements between each infinitesimal entry point; The ΔPcond , i and ΔP evap , i It can be calculated using the following general formula: ΔP i = f ·( L i / D h )·( r 1 · V² / 2) In the formula: f The friction coefficient is calculated. ΔPcond , i Time to take fc ,calculate ΔP evap , i At that time, take fe ; Li For the first i The length of the plate bundle corresponding to each micro-element; D h The equivalent diameter of the plate heat exchanger; V This refers to the flow rate of the refrigerant within the flow channel; e This is a local minimum value, used to prevent the denominator of the ratio term from being zero; λ is the penalty coefficient; ΔT ref This is the reference critical temperature difference.

[0012] The preferred optimization objective function minUJ of the adaptive intelligent controller is as follows: ; In the formula: N The total number of computational steps in the time domain is set to predict the future; α COP This refers to the COP weighting coefficient; COP k Based on the system state prediction model k The state variables are obtained by predicting them in one computational step. β risk This refers to the risk weighting coefficient. c phi This is the dry spot weighting coefficient; r Δu The weighting coefficient for the rate of change of the control quantity; Thu k It is the difference between the control decision variables of two adjacent computation steps in the adaptive intelligent controller optimization problem; The soft constraint quadratic penalty function is defined as follows: The R threshold The threshold for the risk index of heat transfer deterioration; The f max This is the maximum permissible value for dry spot coverage.

[0013] Preferably, the adaptive intelligent controller automatically switches between low-temperature heat source mode, normal mode, and high-temperature heat source mode based on a comparison between the absolute value of the cooling water temperature or the rate of change of the cooling water temperature and a preset threshold for the rate of change of the cooling water temperature. Specifically, the low-temperature heat source mode corresponds to operating conditions where the cooling water temperature is below a first preset temperature, the high-temperature heat source mode corresponds to operating conditions where the cooling water temperature is above a second preset temperature, and the normal mode corresponds to operating conditions between the two. In the low-temperature heat source mode, the dry spot weighting coefficient is increased. c phi Increase the risk weighting coefficient under high-temperature heat source mode β risk In normal mode, the default weights are restored.

[0014] The preferred threshold for the cooling water temperature change rate is based on the total thermal inertia time constant of the cascade high-temperature heat pump. t thermal and mode switching temperature step ΔT step Confirmed, the calculation formula is as follows: ΔT step / (3 t thermal ).

[0015] This invention achieves real-time soft measurement of liquid film thickness and dry spot coverage by constructing a dynamic liquid film observer for a falling film evaporator that includes a liquid film rupture loss term; it predicts the core operating state of the cascade high-temperature heat pump within a future set time domain by constructing a system dynamic prediction model based on liquid film thickness, dry spot coverage, and the operating state of the cascade high-temperature heat pump; it quantifies pinch point risk by establishing a micro-element friction-through heat transfer model for a plate-and-shell heat exchanger; and it employs multi-objective rolling optimization with soft constraints and secondary penalties for collaborative control, deriving multi-mode adaptive switching thresholds based on system thermal inertia. Compared with existing technologies, the technical effects of this invention are specifically reflected in the following points: 1. Operational security has been significantly improved. The dry spot coverage rate is reduced from more than 10% in existing technologies to less than 5%, completely eliminating the risk of compressor liquid slugging; the minimum temperature difference of the plate heat exchanger is stably maintained above 1.5K, preventing local heat transfer deterioration; the frequency of actuator operation is reduced by 40%, reducing equipment start-up and shutdown wear, and doubling the equipment maintenance cycle.

[0016] 2. Energy efficiency has been significantly improved. The system's COP is 8-12% higher than that of conventional PID control, and the 960kW unit saves about 400,000 kWh of electricity per year, reducing long-term energy costs from the source.

[0017] 3. Optimization of control accuracy and adaptability to operating conditions The control accuracy is improved by 40% compared with the traditional solution, and there is no oscillation overshoot problem; it can adapt to the cooling water temperature change of 20~40℃, the mode switching is shockless, the outlet water temperature fluctuation is less than or equal to 0.4℃, and the 50ms PID inner loop ensures the response speed of the actuator, the outer loop is optimized in 10 seconds to complete the full working condition calculation, and the single-step solution time is less than 25ms.

[0018] 4. Strong adaptability to industrial applications It adopts a hierarchical control architecture, which can be directly ported to mainstream industrial PLCs such as Siemens S7-1200 / 1500 without hardware upgrades; the payback period for retrofitting existing equipment is less than 1 year, and the project implementation is economical and compatible. Attached Figure Description

[0019] Figure 1 This is a block diagram of the collaborative control architecture according to an embodiment of the present invention.

[0020] Figure 2 This is a block diagram illustrating the principle of the liquid film dynamic observer according to an embodiment of the present invention. Detailed Implementation Example 1

[0021] The adaptive intelligent control method for cascade high-temperature heat pumps of the present invention is as follows: Figure 1As shown, an adaptive intelligent controller, a PID controller, and a cascaded high-temperature heat pump consisting of a low-temperature stage and a high-temperature stage are provided. The PID controller has an inner and outer loop cascade structure. A liquid film dynamic observer for the falling film evaporator is established to achieve real-time soft measurement of liquid film thickness and dry spot coverage. A system dynamic prediction model is constructed, and based on real-time soft measurement data and the operating parameters of the cascaded high-temperature heat pump, the core operating state of the cascaded high-temperature heat pump within a future set time domain is predicted. A micro-element friction heat transfer model of the plate-shell heat exchanger is established to quantify the pinch risk during operation. The adaptive intelligent controller receives the core operating state output by the system dynamic prediction model and the pinch risk data quantified by the micro-element friction heat transfer model of the plate-shell heat exchanger. Based on multi-objective rolling optimization with soft constraint secondary penalty, a dynamic setpoint is generated and sent to the outer loop of the PID controller. The outer loop of the PID controller outputs a control quantity according to the dynamic setpoint to regulate the frequency of the low-temperature stage compressor, the frequency of the high-temperature stage compressor, and the opening of the high-temperature stage electronic expansion valve. The inner loop of the PID controller tracks the dynamic setpoint of the outer loop of the PID controller in real time and controls the opening of the electronic expansion valve of the low-temperature stage.

[0022] The cascade high-temperature heat pump includes a plate-and-shell evaporator-condenser connecting a low-temperature stage and a high-temperature stage, wherein the low-temperature stage adopts a falling film evaporator, and the control method is implemented according to the following specific steps: S1. The liquid film dynamic observer of the falling film evaporator first collects the operating parameters of the low-temperature stage, and solves the liquid film state prediction value through dynamic equations; then, it solves the superheat deviation and liquid film rupture loss based on the liquid film state prediction value; then, it corrects the liquid film state prediction value based on the superheat deviation and liquid film rupture loss; and iterates to obtain the real-time estimation result; the superheat deviation is the difference between the measured suction superheat and the predicted suction superheat. S2. The system dynamic prediction model takes the current core operating state and liquid film state of the cascade high-temperature heat pump as internal state variables, receives the control commands output by the adaptive intelligent controller as input vectors, and calculates and outputs the core operating state parameters of the cascade high-temperature heat pump in the future set time domain. S3. The friction heat transfer model of the plate-shell evaporator-condenser divides the plate bundle into M micro-elements along the working fluid flow direction. Based on the output of the system dynamic prediction model, the local heat transfer temperature difference between the condensing side and the evaporating side in each micro-element is solved. Then, the ratio of the maximum to the minimum local temperature difference along the plate bundle is calculated. Combined with the exponential penalty term of the minimum local temperature difference, the heat transfer deterioration risk index is obtained. S4. The adaptive intelligent controller is constructed based on the output values ​​of the system dynamic prediction model and the friction heat transfer model of the plate-and-shell evaporator-condenser. It includes a system performance coefficient COP maximization term, a heat transfer deterioration risk index penalty term, a dry spot coverage penalty term, and a regularization term. The heat transfer deterioration risk index penalty term and the dry spot coverage penalty term both adopt a soft-constraint quadratic penalty form, and the penalty is activated only when the risk index or dry spot coverage exceeds its respective threshold.

[0023] The embodiment of this invention uses a 960kW R1234ze(E) / R1233zd(E) cascade high-temperature heat pump system, which is applied to a waste heat recovery project of cooling water in a brewery. The system uses 35°C cooling water as a low-grade heat source to produce 90°C high-temperature hot water for beer sterilization.

[0024] The system hardware configuration is as follows: Low-temperature stage cycle: semi-hermetic screw compressor, displacement 930m³ / h, rated power 200kW; falling film evaporator, heat exchange area 25m², double-sided reinforced high-efficiency copper heat exchange tubes with specifications Φ19.05×1.04mm; electronic expansion valve with a diameter of 24mm.

[0025] High-temperature stage cycle: semi-hermetic screw compressor, displacement 930m³ / h, rated power 132kW; high-temperature stage shell and tube condenser, heat exchange area 35m², double-sided reinforced high-efficiency copper heat exchange tubes with specifications φ19.05×0.95mm.

[0026] Coupled heat exchanger: Plate-shell evaporator-condenser with a heat exchange area of ​​80m² and a plate spacing of 2.5mm.

[0027] Control system: The adaptive intelligent controller is a Siemens S7~1215C PLC, which adopts a hierarchical control architecture: the inner loop of 50ms PID control controls the low-temperature electronic expansion valve, and the outer loop of 10 seconds optimizes the calculation of multiple objectives such as the frequency of the high-temperature electronic expansion valve and the high and low temperature compressor; equipped with high-precision pressure sensors (range 0~2.5MPa, accuracy ±0.5%FS), temperature sensors (range -50~150℃, accuracy ±0.1℃), and electromagnetic flowmeters (range 0~200m³ / h, accuracy ±0.5%FS).

[0028] The operation flow of each control cycle in Embodiment 1 of the present invention is as follows: System initialization: After the unit is powered on, a self-test procedure is executed to initialize parameters such as the initial threshold and iteration step size of the liquid film state observer, as well as core parameters such as the weight coefficient and mode switching threshold of the adaptive intelligent controller; Automatic parameter identification: After the unit has preheated and stabilized, the adaptive intelligent controller automatically applies the fracture coefficient. k break The online identification process took approximately 5 minutes to complete the initialization of core parameters.

[0029] During normal operation: A dual-layer control cycle is adopted. First, the liquid film state of the heat exchange tube is acquired in real time through the liquid film dynamic observer and the future state is deduced by the system dynamic prediction model. Then, the local temperature difference and the risk of heat transfer deterioration are calculated by the friction heat transfer model of the plate heat exchanger. The adaptive controller collects parameters every 10 seconds and solves the constrained multi-objective optimization problem, adjusts the weight coefficients of the optimization objectives, outputs the optimal control sequence, updates the PID outer loop setpoint, and regulates the frequency of the high and low temperature stage compressor and the opening of the high temperature stage electronic expansion valve. At the same time, the PID inner loop tracks the dynamic setpoint of the PID outer loop with a period of 50ms to quickly adjust the opening of the low temperature stage electronic expansion valve.

[0030] Anomaly protection employs a two-level mechanism: optimizing the upper limit of dry spot coverage φ in the constraints. max =5% (normal soft constraint). When the dry spot coverage exceeds 8% or the heat transfer deterioration risk index exceeds 10, the secondary safety protection (forced frequency reduction) is triggered. When a sensor failure or communication abnormality occurs, the system switches to PID backup control mode.

[0031] The liquid film dynamic observation device for the falling film evaporator is constructed based on the falling film evaporation heat transfer-flow coupling mechanism, and its principle block diagram is as follows: Figure 2 As shown: State prediction equations for average liquid film thickness and dry spot coverage; The state prediction equation for the average thickness of the liquid film is as follows: ; The state prediction equation for the dry spot coverage is as follows: when hour, ; when d ≥ crit hour, ; In the formula: dδ / dt The rate of change of the average thickness of the liquid film, in meters per second; dφ / dt The rate of change of dry spot coverage of the liquid film is dimensionless. d This is an estimate of the average thickness of the liquid film, in meters. d 0 The initial liquid film thickness, ,in The rated refrigerant spray flow rate is expressed in kilograms per second. f This is an estimated value for the dry spot coverage of the liquid film, dimensionless, with an initial value of zero; The refrigerant spray mass flow rate at the outlet of the low-temperature electronic expansion valve is expressed in kilograms per second. The mass flow rate of refrigerant evaporation in the evaporator, in kilograms per second. , Heat exchange in the evaporator, unit: kilowatt; r The latent heat of vaporization of refrigerant R1234ze(E); Mass flow rate lost due to liquid film rupture, unit: kg / s; ; In the formula K break The liquid film rupture loss coefficient is initially set at 0.15. Its subsequent value is directly related to the orifice ratio of the distributor, the tube spacing, and the surface tension of R1234ze(E) in the falling film evaporator. Online identification and automatic calibration ensure that the estimated rupture loss conforms to the actual physical process. d ≥ d crit At that time, the burst loss term is 0; Parameter automated online identification: After the unit preheats and stabilizes, the adaptive intelligent controller gradually reduces the opening of the low-temperature stage electronic expansion valve according to a preset step size, while simultaneously collecting the pressure drop change rate at the evaporator inlet and outlet in real time; when the pressure drop change rate reaches an inflection point, the current refrigerant flow rate and estimated liquid film thickness are recorded and substituted sequentially into... dδ / dt, Calculation formula reverse calculation K break And automatically update to the adaptive intelligent controller storage area; , d crit The critical liquid film thickness is determined by taking the initial spray liquid film thickness. d 0 25%; Density of refrigerant R1234ze(E), unit: kg / m³; The total wetted circumferential area of ​​the evaporator tube bank, in square meters; ; in W peri For the wetting perimeter of a single tube, N tube For the total number of people under management, L tube For effective management; , The observer gain is determined using the pole placement method, placing the observer poles at 3 to 5 times the real part of the system poles to ensure that the observer convergence speed is much faster than the system dynamics. In this embodiment, the thermal inertia time constant τ = 180 s, and the observer time constant is taken as 60 s. Take the negative value. Take positive values, unit: meters / (second·Kelvin). The value ranges from -0.0005 to -0.002 m / (s·K), and in this embodiment 1, the value is -0.001 m / (s·K). The value ranges from 0.3 to 0.8 / (s·K), and in this embodiment 1, it is 0.5 / (s·K). This is the measured value of the suction superheat of the cryogenic compressor, in Kelvin. The compressor suction superheat is predicted based on the liquid film state, in Kelvin. The Based on the current moment d and f Estimated value, effective heat transfer conductivity modulated by liquid film Ueff(δ,φ) Establish a mapping relationship between the liquid film state and superheat prediction to form a closed loop for the output error feedback of the observer. The specific formula is as follows: ; In the formula: K sh The superheat conversion coefficient reflects the proportion of the average heat transfer temperature difference converted into outlet steam superheat. It is dimensionless and ranges from 0.3 to 0.6. In this example 1, the value is 0.4. Ueff(δ,φ) · A evap Effective heat transfer conductivity modulated by the liquid film, unit: kW / Kelvin; ; In the formula U wet (d) The heat transfer coefficient of the wetted region is expressed in a simplified form using the Nusselt vertical falling film heat transfer coefficient: U wet (d) = C·(n / n 0 ) (-1 / 3) , C The reference heat transfer coefficient is expressed in watts per square meter (Kelvin). For R1234ze(E), the value is 2.5 × 10³ W / (m²·K). U dryThe heat transfer coefficient of the dry spot region is taken as a value. U wet (d) 20%; ΔT sh ,0 This is a baseline superheat correction term used to correct superheat loss caused by pipe wall thermal resistance and flow non-uniformity. Its value ranges from 1 to 2K, and in this embodiment 1, it is 1.5K. d crit The critical liquid film thickness is defined as a value of [value to be filled in]. d 0 25%; α dry The value is the dry spot spread rate coefficient, ranging from 0.1 to 0.5 / s, and is 0.2 / s in this embodiment 1. -α recover The value is the dry spot recovery rate coefficient, ranging from 0.05 to 0.2 / s. In this example 1, it is 0.1 / s.

[0032] The system dynamic prediction model is as follows: dx / dt = A · x + B · u ; In the formula: x For state variables, x = [ P evap , P cond , T mid , d , f ] T The P evap The evaporation pressure at the low temperature stage. P cond For high-temperature condensation pressure, T mid The intermediate temperature; u For the input vector, u = [ f low , f high , v high ] T The f low For the frequency of the cryogenic compressor, f high For high-temperature compressor frequency,v high This refers to the opening degree of the high-temperature electronic expansion valve. A is a 5×5 diagonally dominant matrix, where the diagonal elements reflect the thermal inertia of each component, with a typical value of a. 11 =-1 / t evap a 22 = -1 / t cond a 33 = -1 / t mid The t evap The thermal inertia time constant of the low-temperature stage evaporator, the t cond The thermal inertia time constant of the high-temperature stage condenser, the t mid The time constant of thermal inertia for the intermediate temperature range. t cond ,t cond and t mid The units are all seconds; off-diagonal elements reflect T mid and P evap and P cond The coupling relationship; B is the input matrix, which reflects the adjustment gain of the input vector on the state variables; Matrix A and B are obtained through the following standardized steps: After the unit is running stably under rated conditions, a ±5% step disturbance is applied to the frequency of the cryogenic compressor, the frequency of the high-temperature compressor, and the opening of the electronic expansion valve, respectively. Pressure and temperature response data are recorded with a 1-second sampling period, and the discretized state-space model parameters are identified using the least squares method. The step response identification is performed under the condition that the PID controller parameters of the cryogenic electronic expansion valve have been tuned and locked according to process requirements. The identified generalized controlled object model includes the dynamic characteristics of the fixed PID. If the PID parameters change subsequently, the identification needs to be performed again.

[0033] The heat transfer deterioration risk index Rk The calculation formula is as follows: Rk = max( ΔTi,k ) / max[min( ΔTi,k ), e ] + λ·exp[-min (ΔTi,k) / ΔT ref ]; In the formula: max( ΔTi,k The actual calculated values ​​are used without ε truncation; ε is a minimum value (e.g., 0.1K), used only to prevent numerical solution failure due to a zero denominator in the ratio term; the actual minimum temperature difference min is retained in the exponential penalty term. (D Ti,k) This ensures that the exponential term effectively increases when the pinch deteriorates, and together with the ratio term, constitutes a non-smooth, strong penalty effect. k This indicates the sequence number of the computation step within the specified time domain for predicting the future. ΔTi For the first i Local heat transfer temperature difference within a micro-element, unit: Kelvin; ; In the formula: T sat This is a saturation temperature function, representing the saturation temperature of the refrigerant at a given pressure, in Kelvin. P cond , in The condensing pressure at the inlet of the low-temperature stage condenser, in Pa or bar; P evap , in The evaporation pressure at the inlet of the high-temperature evaporator, in Pa or bar; SDPcond , i To the condenser side inlet to the first i The sum of the cumulative pressure drops of all infinitesimal elements between each infinitesimal entry point; Σ ΔP evap , i To the evaporator side inlet to the first i The sum of the cumulative pressure drops of all infinitesimal elements between each infinitesimal entry point; The ΔPcond , i and ΔP evap , i It can be calculated using the following general formula: ΔP i = f ·( L i / D h )·( r 1 · V² / 2) In the formula: f The friction factor is dimensionless; calculation ΔPcond , i Time to take fc That is, R1234ze(E) condenser-side resistance coefficient, with a value of 0.025; Calculate ΔP evap , i At that time, take oh, That is, R1233zd(E) is the evaporator-side drag coefficient, with a value of 0.03; Li For the first i The length of the plate bundle corresponding to each micro-element, in meters; D h The equivalent diameter of the plate heat exchanger, in meters; V The velocity of the refrigerant within the flow channel, measured in meters per second. λ is the penalty coefficient, with a value of 5.0; ΔT ref The reference critical temperature difference is set to 1.5K; The optimization objective function minUJ of the adaptive intelligent controller is as follows: ; In the formula: N To predict the total number of computation steps in the future time domain, the value ranges from 15 to 25, and in this embodiment 1, it is set to 20; α COP The COP weighting coefficient is 0.75 in this embodiment 1. COPk Based on the system state prediction model k The state variables predicted in each calculation step are obtained by mapping these state variables to power and heat through steady-state algebraic relationships (compressor performance curves + heat exchanger heat transfer equations). The calculation formula is as follows: COP = Q condensing / ( W low + W high ) In the formula, Q condensing Provides heating for the high-temperature stage condenser; W low and W high These are the electrical power ratings for the low-temperature stage and the high-temperature stage compressors, respectively.

[0034] β riskThe risk weighting coefficient is set to 1 in this embodiment 1; c phi The dry spot weighting coefficient is set to 25 in this embodiment 1; r Δu The weighting coefficient for the rate of change of the control quantity is set to a value ranging from 0.05 to 0.2; in this embodiment 1, it is set to 0.1. Thu k It is the difference between the control decision variables in two adjacent computation steps in the adaptive intelligent controller optimization problem. u k-1 Based on historical execution values, u k These are the decision variables to be optimized. Both are obtained directly within the controller and do not depend on external model predictions. The soft constraint quadratic penalty function is defined as follows: The R threshold The threshold for the heat transfer deterioration risk index is set at 3.0, which is dimensionless and leaves a margin between it and the secondary protection trigger index of 10. The f max This is the maximum allowable value for dry spot coverage, which is set to 5% in this embodiment 1.

[0035] Example 2: Similar to Example 1, the difference is that the adaptive intelligent controller automatically switches between low-temperature heat source mode, normal mode, and high-temperature heat source mode based on the comparison between the absolute value of the cooling water temperature or the rate of change of the cooling water temperature (i.e., the first derivative of the cooling water temperature) and a preset threshold for the rate of change of the cooling water temperature. Specifically, the low-temperature heat source mode corresponds to the operating condition where the cooling water temperature is lower than a first preset temperature, the high-temperature heat source mode corresponds to the operating condition where the cooling water temperature is higher than a second preset temperature, and the normal mode corresponds to the operating condition between the two. In the low-temperature heat source mode, the dry spot weighting coefficient is increased. c phi Increase the risk weighting coefficient under high-temperature heat source mode β risk In normal mode, the default weights are restored, that is: If the cooling water temperature is <25℃ or the cooling water temperature change rate is <a negative cooling water temperature change rate threshold, then switch to low-temperature heat source mode. The control priority is liquid film stability > COP > pinch point safety, with corresponding weights of 1 / 2. c phi =30, α COP =0.7, βrisk =10; If the cooling water temperature is 25~33℃ or the absolute value of the cooling water temperature change rate is less than or equal to the threshold value, switch to normal mode with COP priority. When the liquid film exceeds the limit, pinch point safety takes priority, with corresponding weights as follows: c phi =25, α COP =0.75, β risk =12; If the cooling water temperature is greater than 33℃ or the cooling water temperature change rate is greater than the cooling water temperature change rate threshold, then switch to high-temperature heat source mode. The control priority is pinch point safety > COP > liquid film stability, with corresponding weights of [missing information]. c phi =20, α COP =0.8, β risk =15; The threshold for the cooling water temperature change rate is based on the total thermal inertia time constant of the cascade high-temperature heat pump. t thermal and mode switching temperature step ΔT step Confirmed, the calculation formula is as follows: ΔT step / (3 t thermal ),in ΔT step The value ranges from 4 to 6℃, and in this example 2, 5℃ is used. t thermal The value range is 2 to 5 minutes. In this embodiment 2, 3 minutes is used. For a typical configuration with τ_thermal = 3 minutes, the threshold for the cooling water temperature change rate is about 0.56℃ / min. In engineering applications, 0.5℃ / min can be used as an approximation.

[0036] Comparison of effects and results:

[0037] Comparative example: The existing PID control scheme is adopted, and the parameters have been tuned using the Ziegler-Nichols method, achieving optimal steady-state performance under rated operating conditions.

[0038] The comparative examples and Embodiments 1 and 2 of the present invention were operated under the same rated conditions (cooling water inlet temperature 35°C and hot water outlet temperature 90°C).

[0039] The COP comparison results of the comparative example and the different working conditions of Example 1 of the present invention are shown in Table 1.

[0040] Table 1

[0041] The results show that under rated steady-state conditions, the COP of Embodiment 1 of the present invention is improved by 11% compared with PID control; within the steady-state operating range of cooling water temperature from 20 to 40℃, the COP is improved by 8% to 12%; under dynamic disturbance conditions with continuous step changes in cooling water temperature (25℃→20℃→35℃→40℃), the average COP in the time domain is improved by about 5%, while the control stability is improved by 2.1 times.

[0042] The results of comparing the dry spot coverage of the comparative example with that of Example 1 of the present invention are shown in Table 2.

[0043] Table 2

[0044] The results showed that the dry spot coverage of the comparative example varied greatly, especially in the low-temperature section of cooling water at 20°C, where the dry spot coverage was >12%, which was seriously excessive; while the dry spot coverage of Example 1 of the present invention was stable between 2.5% and 3.5%.

[0045] The results of the outlet water temperature change during the multi-mode switching process of the comparative example and Embodiment 2 of the present invention, namely, the cooling water temperature jumps from 25°C to 20°C (low temperature heat source mode), then jumps to 35°C (conventional mode), and then rises to 40°C (high temperature heat source mode), are shown in Table 3.

[0046] Table 3

[0047] The results show that: Embodiment 2 of the present invention achieves shock-free mode switching and keeps the outlet water temperature fluctuation within ±0.4℃ by dynamically adjusting the mode switching threshold and multi-objective weights based on thermal inertia; while the comparative example shows obvious overshoot and long-term oscillation after each cooling water temperature jump.

[0048] In summary, the COP of this invention reaches 3.1, an improvement of 11% compared to existing PID control schemes; the dry spot coverage is stable between 2.5% and 3.5%, with no heat transfer deterioration caused by dry spots; the minimum temperature difference of the plate heat exchanger is maintained between 1.7 and 1.9 K, with no pinch points; the hot water outlet temperature fluctuation is ≤ ±0.4℃, fully meeting the process requirements; when the cooling water temperature changes within the range of 20 to 40℃, the system automatically switches operating modes without impact or oscillation; the average single-step calculation time for optimization on the Siemens S7-1215CPLC is <25ms, fully meeting the real-time requirement of a 10-second control cycle.

Claims

1. An adaptive intelligent control method for a cascade high-temperature heat pump, comprising an adaptive intelligent controller, a PID controller, and a cascade high-temperature heat pump consisting of a low-temperature stage and a high-temperature stage, wherein the PID controller has an inner and outer loop cascade structure, characterized in that: A liquid film dynamic observer for a falling film evaporator is established to achieve real-time soft measurement of liquid film thickness and dry spot coverage. A system dynamic prediction model is constructed, and based on real-time soft measurement data and operating parameters of the cascade high-temperature heat pump, the core operating state of the cascade high-temperature heat pump within a future set time domain is predicted. A micro-element friction heat transfer model for a plate-shell heat exchanger is established to quantify pinch risk during operation. The adaptive intelligent controller receives the core operating state output by the system dynamic prediction model and the pinch risk data quantified by the micro-element friction heat transfer model of the plate-shell heat exchanger, and generates a dynamic setpoint based on multi-objective rolling optimization with soft constraint secondary penalty and sends it to the outer loop of the PID controller. The outer PID loop outputs a control quantity based on the dynamic setpoint to regulate the frequency of the cryogenic compressor, the frequency of the high-temperature compressor, and the opening of the high-temperature electronic expansion valve; the inner PID loop tracks the dynamic setpoint of the outer PID loop in real time to control the opening of the cryogenic electronic expansion valve.

2. The adaptive intelligent control method for cascade high-temperature heat pumps according to claim 1, characterized in that: The cascade high-temperature heat pump includes a plate-and-shell evaporator-condenser connecting a low-temperature stage and a high-temperature stage, wherein the low-temperature stage uses a falling film evaporator, and the following specific steps are performed: S1. The liquid film dynamic observer of the falling film evaporator first collects the operating parameters of the low-temperature stage, and solves the liquid film state prediction value through dynamic equations; then, it solves the superheat deviation and liquid film rupture loss based on the liquid film state prediction value; then, it corrects the liquid film state prediction value based on the superheat deviation and liquid film rupture loss; and iterates to obtain the real-time estimation result; the superheat deviation is the difference between the measured suction superheat and the predicted suction superheat. S2. The system dynamic prediction model takes the current core operating state and liquid film state of the cascade high-temperature heat pump as internal state variables, receives the control commands output by the adaptive intelligent controller as input vectors, and calculates and outputs the core operating state parameters of the cascade high-temperature heat pump in the future set time domain. S3. The friction heat transfer model of the plate-shell evaporator-condenser divides the plate bundle into M micro-elements along the working fluid flow direction. Based on the output of the system dynamic prediction model, the local heat transfer temperature difference between the condensing side and the evaporating side in each micro-element is solved. Then, the ratio of the maximum to the minimum local temperature difference along the plate bundle is calculated. Combined with the exponential penalty term of the minimum local temperature difference, the heat transfer deterioration risk index is obtained. S4. The adaptive intelligent controller is constructed based on the output values ​​of the system dynamic prediction model and the friction heat transfer model of the plate-and-shell evaporator-condenser. It includes a system performance coefficient COP maximization term, a heat transfer deterioration risk index penalty term, a dry spot coverage penalty term, and a regularization term. The heat transfer deterioration risk index penalty term and the dry spot coverage penalty term both adopt a soft-constraint quadratic penalty form, and the penalty is activated only when the risk index or dry spot coverage exceeds its respective threshold.

3. The adaptive intelligent control method for cascade high-temperature heat pumps according to claim 2, characterized in that: The liquid film dynamic observer of the falling film evaporator includes a state prediction equation for the average liquid film thickness and a state prediction equation for the dry spot coverage. The state prediction equation for the average thickness of the liquid film is as follows: ； The state prediction equation for the dry spot coverage is as follows: when δ < δcrit hour, ; when δ ≥ δcrit hour, ; In the formula: dδ / dt This represents the rate of change of the average thickness of the liquid film. dφ / dt The rate of change of the dry spot coverage of the liquid film; δ This is an estimated average thickness of the liquid film. φ This is an estimated value for the dry spot coverage of the liquid film; The refrigerant spray mass flow rate at the outlet of the low-temperature electronic expansion valve; The mass flow rate of refrigerant evaporation within the evaporator is... , For heat exchange in the evaporator, r The latent heat of vaporization of the refrigerant; The loss of mass flow rate is due to liquid film rupture; ； In the formula K break The liquid film rupture loss coefficient; g(·) = [max(0, ( δ crit -δ ) / δ crit )]²; δ crit The critical liquid film thickness is determined by taking the initial spray liquid film thickness. δ 0 20-30%; The density of the refrigerant; This refers to the total circumferential wetted area of ​​the evaporator tube bank. ; in W peri For the wetting perimeter of a single tube, N tube For the total number of people under management, L tube For effective pipe length; , For observer gain, Take the negative value. Take positive values; This is the measured value of the suction superheat of the cryogenic compressor; The compressor suction superheat is predicted based on the liquid film state. ； In the formula: K sh The superheat conversion coefficient; Ueff(δ,φ) · A evap Effective heat transfer conductivity modulated by the liquid film; ΔT sh ,0 This is a correction term for the baseline superheat. Ueff(δ,φ)=U wet (δ) ·(1- φ )+ U dry ·φ ； In the formula U wet (δ) The heat transfer coefficient of the wetted region. U wet (δ) = C·(δ / δ 0 ) (-1 / 3) , C As the reference heat transfer coefficient, δ 0 The initial liquid film thickness, ,in , rated This refers to the refrigerant spray flow rate under rated operating conditions. U dry The heat transfer coefficient of the dry spot region is taken as a value. U wet (δ) 10-30%; δ crit The critical liquid film thickness is defined as a value of [value to be filled in]. δ 0 20-30%; α dry This is the coefficient for the rate of dry spot spread; -α recover This represents the dry spot recovery rate coefficient.

4. The adaptive intelligent control method for cascade high-temperature heat pumps according to claim 3, characterized in that: The system dynamic prediction model is as follows: dx / dt = A · x + B · u ； In the formula: x For state variables, x = [ P evap , P cond , T mid , δ , φ ] T The P evap The evaporation pressure at the low temperature stage. P cond For high-temperature condensation pressure, T mid The intermediate temperature; u For the input vector, u = [ f low , f high , v high ] T The f low For the frequency of the cryogenic compressor, f high For high-temperature compressor frequency, v high This refers to the opening degree of the high-temperature electronic expansion valve. A is a 5×5 diagonally dominant matrix, where the diagonal elements reflect the thermal inertia of each component, with a typical value of a. 11 =-1 / τ evap a 22 = -1 / τ cond a 33 = -1 / τ mid The τ evap The thermal inertia time constant of the low-temperature stage evaporator, the τ cond The thermal inertia time constant of the high-temperature stage condenser, the τ mid The time constant of thermal inertia for intermediate temperature stages; off-diagonal elements reflect... T mid and P evap and P cond The coupling relationship; B is the input matrix, which reflects the adjustment gain of the input vector on the state variables.

5. The adaptive intelligent control method for cascade high-temperature heat pumps according to claim 4, characterized in that: [the method is described in the original text]. The heat transfer deterioration risk index is: Rk The calculation formula is as follows: Rk = max( ΔTi,k ) / max[min( ΔTi,k ), ε ] + λ·exp[-min (ΔTi,k) / ΔT ref ]; In the formula: k This indicates the sequence number of the computation step within the specified time domain for predicting the future. ΔT i For the first i Local heat transfer temperature difference within a micro-element; ΔT i = T sat(P cond , in-ΣΔPcond , i ) - T sat ( P evap , in- S ΔP evap , i ); In the formula: T sat This is a saturation temperature function, representing the saturation temperature of the refrigerant at a given pressure; P cond , in This refers to the condensing pressure at the inlet of the low-temperature stage condenser. P evap , in This refers to the evaporation pressure at the inlet of the high-temperature evaporator. ΣΔPcond , i To the condenser side inlet to the first i The sum of the cumulative pressure drops of all infinitesimal elements between each infinitesimal entry point; Σ ΔP evap , i To the evaporator side inlet to the first i The sum of the cumulative pressure drops of all infinitesimal elements between each infinitesimal entry point; The ΔPcond , i and ΔP evap , i It can be calculated using the following general formula: ΔP i = f ·( L i / D h )·( ρ 1 · V² / 2) In the formula: f The friction coefficient is calculated. ΔPcond , i Time to take fc ,calculate ΔP evap , i At that time, take fe ; Li For the first i The length of the plate bundle corresponding to each micro-element; D h This is the equivalent diameter of the plate heat exchanger. V This refers to the flow rate of the refrigerant within the flow channel; ε This is a local minimum value, used to prevent the denominator of the ratio term from being zero; λ is the penalty coefficient; ΔT ref This is the reference critical temperature difference.

6. The adaptive intelligent control method for cascade high-temperature heat pumps according to claim 5, characterized in that: The optimization objective function minUJ of the adaptive intelligent controller is as follows: ； In the formula: N The total number of computational steps in the time domain is set to predict the future; α COP This refers to the COP weighting coefficient; COP k Based on the system state prediction model k The state variables are obtained by predicting them in one computational step. β risk This refers to the risk weighting coefficient. γ phi This is the dry spot weighting coefficient; ρ Δu The weighting coefficient for the rate of change of the control quantity; Δu k It is the difference between the control decision variables of two adjacent computation steps in the adaptive intelligent controller optimization problem; The soft constraint quadratic penalty function is defined as follows: h(R k ) = max(0, R k -R threshold )², the R threshold The threshold for the risk index of heat transfer deterioration; h(φ k ) = max(0, ( (φ k )-φ max ) / φ max )², the φ max This is the maximum permissible value for dry spot coverage.

7. The adaptive intelligent control method for cascade high-temperature heat pumps according to claim 6, characterized in that: The adaptive intelligent controller automatically switches between low-temperature heat source mode, normal mode, and high-temperature heat source mode based on the comparison between the absolute value of the cooling water temperature or the rate of change of the cooling water temperature and a preset threshold for the rate of change of the cooling water temperature. Specifically, the low-temperature heat source mode corresponds to operating conditions where the cooling water temperature is below a first preset temperature, the high-temperature heat source mode corresponds to operating conditions where the cooling water temperature is above a second preset temperature, and the normal mode corresponds to operating conditions between the two. In the low-temperature heat source mode, the dry spot weighting coefficient is increased. γ phi Increase the risk weighting coefficient under high-temperature heat source mode β risk In normal mode, the default weights are restored.

8. The adaptive intelligent control method for cascade high-temperature heat pumps according to claim 7, characterized in that: The threshold for the cooling water temperature change rate is based on the total thermal inertia time constant of the cascade high-temperature heat pump. τ thermal and mode switching temperature step ΔT step Confirmed, the calculation formula is as follows: ΔT step / (3 τ thermal ).

Images