A nonlinear model predictive control method for a two-stage centrifugal refrigeration unit

Abstract

The application discloses a kind of two-stage centrifugal refrigerating unit nonlinear model predictive control methods, by establishing two-stage centrifugal refrigerating unit nonlinear prediction model, the outlet water temperature of evaporator and unit output power are predicted, the objective function of temperature control error, output power and guide vane adjustment increment is constructed, power sensitivity dynamic weight is introduced, and the adjustment increment of high and low pressure stage guide vane is obtained by solving;With the aid of virtual actuator, small amplitude adjustment increment is accumulated, and guide vane actuator dead zone threshold, working condition characteristic index threshold and output interval threshold triple trigger condition are set, only when all conditions are met, guide vane opening control command of next time is output.The defects of existing technology, such as regulation and control lag, actuator dithering wear, insufficient temperature control accuracy and the like, are overcome, the temperature control accuracy of unit is effectively guaranteed, the invalid action of actuator is reduced, the energy consumption is reduced, and the operation stability under variable working condition of unit is improved.

Description

Technical Field

[0001] This invention belongs to the fields of refrigeration and air conditioning equipment control, industrial process control and energy-saving control, and particularly relates to a nonlinear model predictive control method for a two-stage centrifugal chiller unit. Background Technology

[0002] Two-stage centrifugal chiller units typically include low-pressure stage compression, high-pressure stage compression, condensation, throttling, evaporation heat exchange, and chilled water circulation. The cooling capacity and power of the unit are closely related to the opening degree of the high and low pressure stage guide vanes (IGV). By adjusting the opening degree of the high and low pressure stage guide vanes, the unit's inlet air flow, interstage pressure relationship, and evaporation side heat exchange capacity can be changed, thereby achieving the regulation of the chilled water outlet temperature.

[0003] Traditionally, two-stage centrifugal chiller units mostly employ PID control, segmented rule control, or empirical linkage control. When the evaporator outlet water temperature is higher than the set value, the controller increases the opening of the low-pressure stage guide vanes; when the evaporator outlet water temperature is lower than the set value, the controller decreases the opening of the low-pressure stage guide vanes. For the high-pressure stage guide vanes, they are often adjusted synchronously with the low-pressure stage guide vanes at a fixed ratio, or adjusted sequentially according to a preset order. Although this type of control method has a simple structure, it approximates the two-stage compression and heat exchange process as a linear or single-input single-output object, failing to fully consider the coupling relationship between the high and low-pressure stage guide vanes, resulting in the following technical problems: 1. Two-stage centrifugal chiller units have obvious thermal inertia and response lag. After the guide vane opening is changed, the compressor flow rate, evaporator heat exchange and evaporator outlet water temperature will not immediately reach a new equilibrium state. The chilled water circulation time and heat exchange lag will cause the outlet water temperature response to lag by tens of seconds to several minutes. The controller only adjusts the guide vane according to the current temperature deviation, which is prone to over-adjustment, under-adjustment or repeated adjustment near the set value. 2. Guide vane actuators typically have limitations on their working cycle and operating frequency. Taking a 50Hz actuator as an example, if the full stroke (0% to 100%) operation time is 21 seconds and the starting frequency is 70%, then the rest time can be estimated as 21 × (1-70%) / 70% = 9 seconds. The minimum interval corresponding to the full stroke operation is 30 seconds. For partial capacity adjustments, the time can be calculated based on the opening change ratio. For example, when the guide vane opening is adjusted from 65% to 55%, the opening change is 10%, and the corresponding actuator working cycle time is 10% × (21 + 9) = 3 seconds. Considering the chilled water response cycle, longer control action intervals are usually required in engineering, such as about 30 seconds. Therefore, the number of actuator actions per minute is limited to no more than three. Consequently, if the system temperature deviation occurs between two valid output commands of the controller, the controller cannot correct it in time and can only wait for the next control cycle to process it. This results in untimely intervention and reduced system control accuracy. 3. When the guide vane actuator uses a 4-20mA input signal, it also has a dead zone characteristic. If the dead zone or sensitivity of the guide vane actuator is set to 1.8%, the actuator will only operate after the input current change exceeds 16 × 1.8% = 0.288mA. If the small adjustment output by the controller falls within the dead zone of the guide vane actuator, the actual guide vane position will not change. If the controller continuously outputs small commands, it may cause actuator jitter, mechanical wear, and temperature control oscillation.

[0004] To address the problems of the aforementioned control methods, model predictive control (MMC) has been applied to energy-saving control of chiller plants, building air conditioning systems, and chiller unit clusters. Existing nonlinear predictive models based on neural networks use parameters such as unit operating state variables and environmental variables from the current and previous moments as model inputs to predict the unit operating state at the next moment. While it's possible to solve for control variables that satisfy constraints using the model-predicted future state within the prediction time domain, existing MMC methods focus primarily on optimizing external equipment such as chiller plant supply water temperature settings, multiple unit start-up and shutdown, chilled water pumps, cooling water pumps, or cooling tower fans. They only uniformly incorporate the unit actuator duty cycle, maximum number of actions per minute, control dead zone, cumulative adjustment demand within the dead zone, and local sensitivity of the high and low pressure stage guide vanes into the control scope. This approach remains insufficient for the nonlinear coordinated control of the high and low pressure stage guide vanes within a single two-stage centrifugal chiller unit. Summary of the Invention

[0005] The present invention aims to solve the aforementioned technical problems existing in the prior art by providing a nonlinear model predictive control method for a two-stage centrifugal chiller unit.

[0006] The technical solution of this invention is: a nonlinear model predictive control method for a two-stage centrifugal chiller unit, which executes the following steps in sequence: Step 1. Collect the current operating data of the two-stage centrifugal chiller unit; Step 2. Preprocess the collected data to obtain the known unit state vector at the current moment; Step 3. Based on the known unit state vector, use the nonlinear prediction model of the two-stage centrifugal chiller unit to predict the unit controlled parameters at the next moment; Step 4. Perform constrained rolling optimization based on the objective function to obtain the next time step of the guide vane opening adjustment increment of the high and low pressure stages; Step 5. Determine whether the next moment's high and low pressure stage guide vane opening adjustment increment is greater than the dead zone threshold. If no, proceed to step 6. If yes, assign the next moment's high and low pressure stage guide vane opening adjustment increment to the cumulative adjustment amount and proceed to step 7. Step 6. Input the next moment's high and low pressure stage guide vane opening adjustment increment into the virtual actuator for accumulation, obtain the next moment's cumulative adjustment amount, and then determine whether the next moment's cumulative adjustment amount is greater than the dead zone threshold. If yes, proceed to step 7; otherwise, return to step 1. Step 7. Determine whether the current operating condition characteristic index is greater than the operating condition characteristic index threshold. If no, proceed to step 8. If yes, return to step 1. Step 8. Determine whether the timing duration between two valid output control commands is greater than the output interval threshold; if no, return to step 1; if yes, add the actual opening of the high and low pressure stage guide vanes at the current moment to the cumulative adjustment amount at the next moment to obtain the high and low pressure stage guide vane opening control command at the next moment and output it; clear the cumulative adjustment amount of the virtual actuator to zero and return to step 1.

[0007] The preferred objective function As shown in the following formula: ;

[0008] in, To predict the length of the time domain; Set the evaporator outlet water temperature; k The current time number; For the prediction step sequence number; for Predict the evaporator outlet water temperature at any time; for Predicted unit power values ​​at any given time; , , These are the weighting coefficients for temperature tracking, power control, and guide vane opening adjustment, respectively. It is a high-voltage stage; It is a low-pressure stage; for The increment of the high-pressure stage guide vane opening adjustment at any given moment; for The increment of the low-pressure stage guide vane opening at any given moment; The optimization solution of the objective function should satisfy the constraints. The Dynamically assign weights to the high-pressure stage guide vanes. The dynamic weighting of the low-pressure stage guide vanes is calculated according to the following formula: ; ; ;

[0009] in, Index for high-voltage or low-voltage stage; for ktime Dynamic weight allocation for guide vanes; This is the power cost factor; To sum the indices; for k time The local sensitivity of guide vanes to predicting evaporator outlet water temperature; for k time The local sensitivity of the guide vanes to the predicted unit power; for Guide vanes are used for small perturbations in local differential operations; for k time The actual opening of the guide vane; for k- 1. Measurement value of evaporator outlet water temperature of the unit at time 1; for k- Power measurement of the unit at time 1.

[0010] Preferred operating condition characteristic indicators at the current time Calculate according to the following formula: ; in, for k The measured value of the evaporator outlet water temperature of the unit. for k The rate of change of evaporator outlet water temperature at all times. , , These are the weighting coefficients. and They are respectively k The cumulative adjustment of the high and low pressure guide vanes at any given time.

[0011] Preferred output interval threshold ; ; ;

[0012] in, The interval between actions within the actuator's working cycle; Minimum observation time required for chilled water circulation and temperature response; This represents the maximum number of actions per minute for the guide vane actuator. for k time s The absolute value of the guide vane opening adjustment increment; This refers to the total operating time of the guide vane actuator. This refers to the rest time corresponding to the full stroke of the guide vane actuator; This refers to the startup frequency.

[0013] Compared with existing technical solutions, the present invention has the following advantages: (1) A nonlinear prediction model for a two-stage centrifugal chiller unit is adopted to predict the evaporator outlet water temperature and output power of the unit at future times, providing basic parameters for subsequent objective function optimization.

[0014] (2) The objective function is constructed based on the evaporator outlet water temperature tracking error, the predicted output power of the unit and the action amplitude of the high and low pressure guide vanes. The weights are dynamically allocated based on the sensitivity of the high and low pressure guide vanes to obtain the adjustment increment of each guide vane opening. This achieves coordinated optimization of temperature control accuracy, operating efficiency and action amplitude, and avoids the additional energy loss caused by the fixed proportional linkage of the high and low pressure guide vanes.

[0015] (3) The dead zone threshold, operating condition characteristic index threshold and output interval threshold of the guide vane actuator are used as the trigger conditions for the controller to issue commands. In particular, small candidate adjustment increments below the dead zone threshold of the guide vane actuator are accumulated through the virtual actuator. Only after the trigger conditions are met will the accumulated adjustment amount be used as an effective variable to issue an effective command, so as to avoid the actuator jitter and wear caused by small variable commands and extend the service life of the actuator. Attached Figure Description

[0016] Figure 1 This is a general flowchart of an embodiment of the present invention.

[0017] Figure 2 This is a comparison chart of the outlet water temperature tracking curves of the embodiments of the present invention and the PID control method.

[0018] Figure 3 This is a comparison chart of the total current curves of the embodiments of the present invention and the PID control method. Detailed Implementation

[0019] The present invention provides a nonlinear model predictive control method for a two-stage centrifugal chiller unit, as follows: Figure 1 As shown: Perform the following steps in sequence: Step 1. Data Collection k Real-time operating data of the two-stage centrifugal chiller unit; The operating data of the two-stage centrifugal chiller unit includes the inlet and outlet water temperatures of the evaporator and condenser, the suction pressure of the low-pressure stage, the exhaust pressure of the high-pressure stage, the main motor current of the high and low pressure stages, the guide vane opening control commands of the high and low pressure stages, the actual opening of the guide vanes of the high and low pressure stages, and the liquid levels of the evaporator and condenser. The sampling period is set to 200ms according to the unit response and actuator limitations.

[0020] Step 2. Preprocess the collected data to obtain... k The unit state vector with known time; Preprocessing specifically includes outlier removal, moving average filtering, and state variable normalization. k The known unit state vector includes the real-time data after this preprocessing and k- The unit state vector is stored at time 1. The unit state vector consists of operating state variables, environmental variables, and measured opening of the guide vanes of the high and low pressure stages. The operating state variables include evaporator outlet water temperature, high and low pressure stage main motor current, low pressure stage suction pressure, and high pressure stage exhaust pressure. The environmental variables include evaporator inlet water temperature and condenser inlet water temperature.

[0021] Step 3. Based on the known unit state vector, a nonlinear prediction model for a two-stage centrifugal chiller unit is used to predict... k The unit's controlled parameters at time +1; The nonlinear prediction model is built upon a neural network and is based on... k The model inputs are the constantly known unit operating state variables, environmental variables, and the measured openings of the high- and low-pressure stage guide vanes. k The unit operating status at time +1 is predicted; it can be expressed as: ;

[0022] in, A nonlinear mapping function representing a neural network model; express k The unit's operating status is known at all times; express k Environment variables that are known at all times; for k time Actual guide vane opening; express k The predicted output of the unit at time +1 includes the evaporator outlet water temperature, the main motor current of the high and low pressure stages, the intake pressure of the low pressure stage, and the exhaust pressure of the high pressure stage.

[0023] Step 4. Perform constrained rolling optimization based on the objective function to obtain... k +1 is the increment for adjusting the guide vane opening of the high and low pressure stages; The objective function As shown in the following formula: ;

[0024] in, To predict the length of the time domain; Set the evaporator outlet water temperature; k The current time number; For the prediction step sequence number; for Predict the evaporator outlet water temperature at any time; for Predicted unit power values ​​at any given time; , , These are the weighting coefficients for temperature tracking, power control, and guide vane opening adjustment, respectively. It is a high-voltage stage; It is a low-pressure stage; for The increment of the high-pressure stage guide vane opening adjustment at any given moment; for The increment of the low-pressure stage guide vane opening at any given moment; The optimization solution of the objective function should satisfy the constraints. The Dynamically assign weights to the high-pressure stage guide vanes. The dynamic weighting of the low-pressure stage guide vanes is calculated according to the following formula: ; ; ;

[0025] in, for k time Dynamic weight allocation for guide vanes; This is the power cost factor; To sum the indices; for k time The local sensitivity of guide vanes to predicting evaporator outlet water temperature; for k time The local sensitivity of the guide vanes to the predicted unit power; for Guide vanes are used for small perturbations in local differential operations; for k time The actual opening of the guide vane; for k- 1. Measurement value of evaporator outlet water temperature of the unit at time 1; for k- Power measurement of the unit at time 1.

[0026] In a specific embodiment of the present invention, prediction in the time domain Set to 20; Temperature tracking weight It can be set to 50~150; power weight The guide vane opening adjustment weight can be set from 0.05 to 5. It can be set from 2 to 15. When the outlet water temperature deviation is large, the controller increases the temperature tracking weight. The controller prioritizes restoring the evaporator outlet water temperature; once the outlet water temperature enters the allowable error range, the controller increases the power weight. And guide vane opening adjustment weight This allows the guide vane assembly to gradually converge to a low-power, low-frequency operating state; the pressure ratio range is set to 90-110% of the design allowable pressure ratio, and the upper limit of the motor current or power is set to 90-105% of the rated value.

[0027] The embodiments of the present invention utilize small disturbances in the local differential of the high-voltage stage guide vanes. =2%, small disturbance in the local differential of the low-pressure stage guide vane =3%, which is applied to the high-pressure stage guide vane and the low-pressure stage guide vane respectively. The changes in outlet water temperature and compressor power are calculated using a nonlinear prediction model. When the marginal effect of a certain stage guide vane on the outlet water temperature is stronger and the power cost per unit cooling capacity is lower, the dynamic allocation weight of that stage guide vane is increased. When the action of a certain stage guide vane will lead to a significant increase in power, an increase in pressure ratio, or a decrease in safety margin, its action priority is reduced.

[0028] Step 5. Judgment k At time +1, is the increment of the guide vane opening adjustment for both high and low pressure stages greater than the dead zone threshold? No, proceed to step 6. Yes, proceed to step 6. k +1 instant high and low pressure stage guide vane opening adjustment increment The value is assigned to the cumulative adjustment amount. , represented as ← Proceed to step 7; Dead zone threshold of the present invention This refers to the dead zone threshold of the guide vane actuator, which is determined according to the following steps: When the guide vane uses a 4-20 mA signal, k Actual opening of the guide vane at any given moment With command current The conversion relationship is as follows: ;

[0029] Corresponding current dead zone and guide vane actuator dead zone threshold The relationship between them is: ;

[0030] If the corresponding current dead zone If the current is 0.288mA, then the dead zone threshold of the guide vane actuator is... It is 1.8%.

[0031] Step 6. k +1 instant high and low pressure stage guide vane opening adjustment increment The input is fed into the virtual executor for accumulation, and the result is... k Cumulative adjustment amount at +1 moment , represented as The for k The cumulative adjustment of the guide vane opening at both high and low pressure stages is recorded when the system is first powered on or when the reset trigger condition is met. Then determine the cumulative adjustment amount. Is it greater than the dead zone threshold? If yes, proceed to step 7; otherwise, return to step 1. Step 7. Determine k Time-of-flight operating condition characteristic indicators Is it greater than the threshold (value is 1.0)? If no, proceed to step 8; if yes, return to step 1. The k Time-of-flight operating condition characteristic indicators Calculate according to the following formula: ; in, for k The measured value of the evaporator outlet water temperature of the unit. for k The rate of change of evaporator outlet water temperature at any given time was set at 1.50 ℃ and 2.0 ℃ / min, respectively. , The normalized baseline value; , , These are weighting coefficients, with values ​​of 0.5, 0.5, and 0.25 respectively. and They are respectively k The cumulative adjustment of the high and low pressure guide vanes at any given time.

[0032] Step 8. Determine whether the timing duration between two valid output high and low pressure stage guide vane opening control commands is greater than the output interval threshold. No, return to step 1; Yes, will k Actual opening of high and low pressure stage guide vanes at all times and k Cumulative adjustment amount at time +1 Add them together to get k +1 Time High and Low Pressure Stage Guide Vane Opening Control Command And output, represented as Clear the cumulative adjustment of the virtual actuator to zero and return to step 1; The output interval threshold ; ; ;

[0033] in, The interval between actions within the actuator's working cycle; Minimum observation time required for chilled water circulation and temperature response; This represents the maximum number of actions per minute for the guide vane actuator. for k The absolute value of the guide vane opening adjustment increment at any given moment; This refers to the total operating time of the guide vane actuator. This refers to the rest time corresponding to the full stroke of the guide vane actuator; Startup frequency; In this embodiment of the invention, the guide vane opening is limited to the range of 0~100%, and the preferred effective adjustment range is 15~95%. If the single effective adjustment increment of the high-pressure stage guide vane and the low-pressure stage guide vane is 2~5%, the minimum effective action interval is set to 20 s, and the number of actions of each actuator per minute does not exceed 3.

[0034] Using the control method of this invention, under the operating range of chilled water flow rate at 70-120% of rated flow rate, cooling water inlet temperature at -8℃ to 0℃, and unit load rate at 20-100%, the steady-state deviation of evaporator outlet water temperature after disturbance subsidence can be stabilized within ±0.5-±1.0℃. Compared with conventional PID control or fixed-proportion high and low pressure stage guide vane linkage control, under the same evaporator outlet water temperature and similar load conditions, this invention can reduce the total number of effective actions of high and low pressure stage guide vanes by 10-30%. Through simulation, the outlet water temperature tracking curve and total current curve of this invention embodiment (MPC) and the conventional PID control method are obtained, and the comparison results are as follows: Figure 2 , Figure 3 As shown. From Figure 2 As can be seen from the comparison of the outlet water temperature tracking curves, the steady-state deviation of the embodiment of the present invention is stable within 0.5℃, while from... Figure 3 The comparison of the total current curves shows that, in the common partial load range of 50-90% load rate, the embodiments of the present invention can reduce the average power of the compressor by about 3% and increase the unit COP by 2-5%.

Claims

1. A nonlinear model predictive control method for a two-stage centrifugal chiller unit, characterized in that... Perform the following steps in sequence: Step 1. Collect the current operating data of the two-stage centrifugal chiller unit; Step 2. Preprocess the collected data to obtain the known unit state vector at the current moment; Step 3. Based on the known unit state vector, use the nonlinear prediction model of the two-stage centrifugal chiller unit to predict the unit controlled parameters at the next moment; Step 4. Perform constrained rolling optimization based on the objective function to obtain the next time step of the guide vane opening adjustment increment of the high and low pressure stages; Step 5. Determine whether the next moment's high and low pressure stage guide vane opening adjustment increment is greater than the dead zone threshold. If no, proceed to step 6. If yes, assign the next moment's high and low pressure stage guide vane opening adjustment increment to the cumulative adjustment amount and proceed to step 7. Step 6. Input the next moment's high and low pressure stage guide vane opening adjustment increment into the virtual actuator for accumulation, obtain the next moment's cumulative adjustment amount, and then determine whether the next moment's cumulative adjustment amount is greater than the dead zone threshold. If yes, proceed to step 7; otherwise, return to step 1. Step 7. Determine whether the current operating condition characteristic index is greater than the operating condition characteristic index threshold. If no, proceed to step 8. If yes, return to step 1. Step 8. Determine whether the timing duration between two valid output control commands is greater than the output interval threshold; No, return to step 1; Yes, add the actual opening of the high and low pressure stage guide vanes at the current moment to the cumulative adjustment amount at the next moment to obtain the control command for the opening of the high and low pressure stage guide vanes at the next moment and output it. Clear the cumulative adjustment of the virtual actuator to zero and return to step 1.

2. The nonlinear model predictive control method for a two-stage centrifugal chiller unit according to claim 1, characterized in that... The objective function As shown in the following formula: ； in, To predict the length of the time domain; Set the evaporator outlet water temperature; The current time number; For the prediction step sequence number; for Predict the evaporator outlet water temperature at any time; for Predicted unit power values ​​at any given time; , , These are the weighting coefficients corresponding to temperature tracking, power control, and guide vane opening adjustment, respectively. It is a high-voltage stage; It is a low-pressure stage; for The increment of the high-pressure stage guide vane opening adjustment at any given moment; for The increment of the low-pressure stage guide vane opening at any given moment; The optimization solution of the objective function should satisfy the constraints. The Dynamically assign weights to the high-pressure stage guide vanes. The dynamic weighting of the low-pressure stage guide vanes is calculated according to the following formula: ； ； ； in, Index for high-voltage or low-voltage stage; for time Dynamic weight allocation for guide vanes; This is the power cost factor; To sum the indices; for time The local sensitivity of guide vanes to predicting evaporator outlet water temperature; for time The local sensitivity of the guide vanes to the predicted unit power; for Guide vanes are used for small perturbations in local differential operations; for time The actual opening of the guide vane; for The measured value of the evaporator outlet water temperature of the unit at any given time; for Power measurement values ​​of the generator set at any given time.

3. The nonlinear model predictive control method for a two-stage centrifugal chiller unit according to claim 2, characterized in that... The current working condition characteristic index Calculate according to the following formula: ； in, for The measured value of the evaporator outlet water temperature of the unit. for The rate of change of evaporator outlet water temperature at all times. , , These are the weighting coefficients. and They are respectively The cumulative adjustment of the high and low pressure guide vanes at any given time.

4. The nonlinear model predictive control method for a two-stage centrifugal chiller unit according to claim 3, characterized in that... The output interval threshold ; ； ； in, The interval between actions within the actuator's working cycle; Minimum observation time required for chilled water circulation and temperature response; This represents the maximum number of actions per minute for the guide vane actuator. for time The absolute value of the guide vane opening adjustment increment; This refers to the total operating time of the guide vane actuator. This refers to the rest time corresponding to the full stroke of the guide vane actuator; This refers to the startup frequency.

Images