Uranium enrichment auxiliary process air conditioning box system apc and rto online collaborative method

Abstract

The application belongs to the technical field of automatic control, and specifically discloses a uranium enrichment auxiliary process air conditioning box system APC and RTO online cooperation method, which comprises the following steps: step 1, APC algorithm modeling; step 2, designing a penalty function and an execution cycle T of APC optimization APC ; step 3, designing an RTO optimization penalty function, a weight coefficient and an RTO cycle; and step 4, initializing APC and RTO optimization parameters and starting a continuous cycle operation. According to the method, numerical model and penalty function are constructed, optimization control is realized based on the prediction of the trend of the cascade hall in the future time interval, thus, the disturbance can be prevented and inhibited in a forward-looking manner, the temperature control stability and precision are ensured, and the air conditioning equipment of the auxiliary process system can be operated in an unattended, efficient and energy-saving manner.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of automatic control, and particularly relates to an online coordination method of an air conditioning box system APC and RTO in a uranium enrichment auxiliary process. BACKGROUND

[0002] The air conditioning system of the cascade hall of the second phase of Shaanxi Uranium of China Nuclear adopts direct expansion air conditioning units, each of which is provided with a cold and heat source device. Two air conditioning machine rooms are arranged on the north and south sides of the hall, each of which is provided with 4 sets of a total of 8 sets of air supply and return fan units, which are responsible for temperature and humidity control of the cascade hall of the second phase of Shaanxi Uranium. The direct expansion air conditioning unit mainly consists of an air handling unit (combined air conditioning unit) and a roof machine (outdoor unit) of a wind-cooled heat pump, and the two parts are directly connected by a copper pipe. The evaporator is in the air conditioning box, the refrigerant flows in the evaporator coil, and the air outside the pipe is dehumidified and cooled (or heated) by the evaporator, and then is sent into the room after a series of treatments.

[0003] The air conditioning system of the third phase of Shaanxi Uranium of China Nuclear is provided with 48 air conditioning units, of which 12 units are arranged in the 01d-1 machine room, 12 units are arranged on the north side of the 01d-2 machine room, and a total of 24 units serve the 01d1 cascade hall; 12 units are arranged in the 01d-3 machine room, 12 units are arranged on the south side of the 01d-2 machine room, and a total of 24 units serve the 01d2 cascade hall. The temperature and humidity of the cascade hall are controlled in a group mode, and every 3 units form a group, which is responsible for temperature and humidity control of a part of the hall. The 01d1 hall and the 01d2 hall each have 8 groups. Each group of the system adopts a one-to-two backup operation mode, and the air conditioning unit is composed of an outdoor unit and an indoor unit. The roof compressor of each air conditioning unit uses a set of refrigeration system pipeline, and a total of 4 scroll compressors are provided.

[0004] One of the important indicators of the auxiliary control system of the uranium enrichment process is to ensure the constant stability of the temperature and humidity of the uranium enrichment cascade hall, so as to avoid the fluctuation of the quality of the main process product. According to the production condition requirements of the uranium enrichment process equipment, the temperature and humidity conditions mainly have two aspects of indicators: (1) the dry ball temperature is stably set at the set temperature target value required by the process within ±1℃; (2) the difference between the dry ball temperature and the dew point temperature exceeds the threshold value, so as to prevent dewing and frosting. A plurality of dry ball temperature / relative humidity sensors are installed in the entire cascade hall, and a dew point instrument is installed at the return air inlet of each air conditioning unit, so as to monitor the temperature and humidity conditions of the space distribution in the hall. Since the physical space of the cascade hall is large, the temperature and humidity measuring points are relatively dispersed, and the air conditioning box modules are relatively more, so the temperature and humidity control of the hall environment is a complex problem with multiple inputs and multiple outputs and coupling. Moreover, the fan frequency regulation values of different air supply units (groups) can be different, the number of operating refrigeration compressors of each unit outdoor unit is different, the air supply volume and temperature (corresponding to the refrigeration capacity) of each air conditioning box air supply outlet are also dynamically changed, and the effect on the temperature detection point is also different, so the traditional independent control strategy of each group in the traditional single group lacks cooperation between groups, and there may be energy waste caused by excessive refrigeration or frequent fluctuations of overcooling and overheating. In addition, there is a time delay and different dynamic response time from the air supply outlet to the detection point of the dry ball temperature sensor in the hall, the refrigeration effect of the air conditioning box is different under different outdoor environment temperatures, and the air conditioning box is a typical nonlinear system. In summary, the uranium enrichment air conditioning box is a typical nonlinear, large time delay and strong coupling control system. At present, the start-stop control, temperature adjustment and supporting air conditioning box frequency adjustment of 8 air conditioners in Shaanxi Uranium Phase II and 48 air conditioners in Shaanxi Uranium Phase III are manually operated by manual operation. In the actual operation process, due to the coupling influence of multiple dispersed sensors of multiple air conditioning units, the workload of the operator monitoring and control is large, and human negligence may cause system fluctuation. How to apply closed-loop control of the fan frequency, the refrigeration capacity of the outdoor unit and the cooperative group control of the equipment start-stop has become one of the difficulties in further improving the control accuracy and automation level of the uranium enrichment auxiliary process.

[0005] The air conditioning box has a high proportion of comprehensive energy consumption in the entire uranium enrichment production process. SUMMARY

[0006] The purpose of the present application is to provide a kind of uranium enrichment auxiliary process air conditioning box system multiple target APC and RTO online cooperation method, this method guarantees the control accuracy of hall temperature and humidity, and realizes the unattended operation of auxiliary process system air conditioning equipment, and high-efficiency energy-saving operation.

[0007] The technical scheme for realizing the purpose of the present application is as follows:

[0008] A kind of uranium enrichment auxiliary process air conditioning box system APC and RTO online cooperation method, the method comprises the following steps:

[0009] Step 1, modeling of APC algorithm

[0010] Step 2, designing of penalty function and execution cycle T of APC optimization APC

[0011] Step 3, designing of penalty function, weight coefficient and RTO cycle of RTO optimization

[0012] Step 4, initialization of APC and RTO optimization parameters and starting of continuous cycle operation.

[0013] The step 1 comprises:

[0014] Step 1.1, determining of operation variables MV, controlled variables CV and disturbance variables DV of the system;

[0015] Step 1.2, obtaining of original model group of the air conditioning system according to model identification experiment;

[0016] Step 1.3, model simplification of the air conditioning system according to steady state gain statistics of the original model group;

[0017] Step 1.4, conversion of continuous frequency model into transfer function in discrete space through continuous-to-discrete transformation method.

[0018] The step 1.2 specifically comprises: carrying out corresponding model identification experiment on the CV / MV / DV variables determined in step 1.1; only changing the set value of one MV or DV each time, performing step test, recording real-time data of all CV / MV / DV and saving; continuously performing step test of each MV or DV for multiple times of upward or downward step change of different amplitudes, recording change data of the system after change, applying system identification algorithm, performing p times of model identification on each MV / DV change to obtain (n+m) MV / DV pairs p CV respective array models.

[0019] The step 1.3 specifically comprises: according to the original model group result of step 1.2, statistically obtaining steady state gain of each matrix unit, eliminating the models without gain or the models with small gain to obtain the simplified model group of the air conditioning system.

[0020] The step 2 comprises:

[0021] Step 2.1, designing of penalty function and adjustable parameters of the function of APC;

[0022] Step 2.2, designing of execution cycle and corresponding step length parameters of APC.

[0023] The mathematical expression of the penalty function J(APC) of APC optimization in the step 2.1 is:

[0024] ​

[0025] satisfies

[0026] y low ≤y(k+l|k)≤y high l=1,2,…,H cy

[0027] u low ≤u(k+r-1|k)≤u high r=1,2,…,H cu

[0028] Δu(k+i|k)=u(k+i)-u(k+i-1)

[0029] |Δu(k+i-1|k)|≤Δu limit i=1,2,…,H u

[0030] wherein:

[0031] y low is the lower bound of the output variable, y high is the upper bound of the output variable;

[0032] u low is the lower bound of the input variable, u high is the upper bound of the input variable;

[0033] Δu limit is the maximum rate of change.

[0034] The step 3 comprises:

[0035] Step 3.1 design of RTO penalty function and design of weight parameter;

[0036] Step 3.2 design of execution period of RTO penalty function and design of weight parameter.

[0037] The mathematical expression of the RTO penalty function in the step 3.1 is:

[0038]

[0039] satisfies

[0040]

[0041]

[0042] wherein, J(RTO) is the penalty function of RTO, Ny is the number of hall temperature and humidity measuring points participating in optimization, i represents the sensor subscript;

[0043] t k，iis the average temperature measurement of the i-th measurement point at the k-th call of RTO;

[0044] h k，i is the average humidity measurement of the i-th measurement point at the k-th call of RTO;

[0045] T tgt is the temperature target, tgt is the humidity target value;

[0046] q k，i , r k，i are the weight coefficients of the temperature deviation T and the humidity deviation H respectively;

[0047] T tgt is the lower limit of the allowed temperature target;

[0048] is the upper limit of the allowed temperature target;

[0049] H tgt is the lower limit of the allowed humidity target;

[0050] is the upper limit of the allowed humidity target.

[0051] The step 4 comprises:

[0052] Step 4.1 uses the artificially set T tgt and H tgt initial values to start the APC execution for the APC controller;

[0053] Step 4.2 reads the new feedback value at each T APC time, executes the J(APC) optimization problem solving once, and uses the first step output value to issue to the actuator;

[0054] Step 4.3 enables the optimization solving of J(RTO) at each T RTO time, and then uses the optimized T tgt and H tgt values to update the ys value of the APC controller, and the APC controller always uses the new ys value as the control target before the next RTO optimization period arrives;

[0055] Step 4.4 uses the T tgt and H tgt values updated at the last time to repeat the J(APC) optimization problem solving of Step 4.2 at the first T APC time after T RTO time, and applies the first step optimal output.

[0056] The beneficial technical effects of the present application are:

[0057] 1. Reduce the daily operation volume of auxiliary process systems. Engineering verification shows that this invention can achieve intelligent closed-loop regulation of air-conditioning boxes, with a closed-loop rate of over 90%, reducing daily operation volume by over 80%;

[0058] 2. Significantly improve the temperature and humidity control accuracy of the cascade hall. Open-loop and closed-loop comparative tests have shown that temperature fluctuations can be reduced from the original + / -1.5 degrees to + / -1 degree. Control performance has been improved by more than 30%;

[0059] 3. An RTO optimization strategy has been designed to dynamically optimize the operating conditions of the air-conditioning box based on the historical operating conditions within the RTO optimization period and the ambient temperature and humidity of the cascade hall. The optimal target value for the best operating condition of the air-conditioning box within the next RTO optimization period is obtained and sent to the real-time controller as the APC control target.

[0060] 4. An APC optimization strategy was designed. The numerical model of the air supply system response and dew point temperature response of the air-conditioning box was used to predict temperature changes in the future time period. The optimal operating target was output based on the RTO rolling optimization, and combined with the outdoor environmental temperature and humidity disturbances. The quadratic programming problem was solved or the optimal control quantity change trajectory was obtained. The first step control quantity was used to adjust the compressor load and the fan frequency of the air-conditioning box equipment, thereby achieving real-time optimization control and energy-saving operation of multiple air-conditioning boxes.

[0061] 5. The development and application of an online collaborative optimization strategy for APC and RTO in the auxiliary process air conditioning system, through the construction of numerical models and penalty functions, enables optimized control based on the prediction of future trends in the cascade hall. This proactively prevents and suppresses disturbances, ensuring stable and precise temperature control. Compared to existing traditional feedback-based closed-loop control or manual operator intervention strategies based on observed fluctuations, this predictive approach addresses the frequent adjustments and overshoots that can result from real-time feedback control. This strategy is of strategic significance for the independent development and innovative breakthroughs of intelligent control technology for auxiliary process air conditioning systems in the uranium enrichment industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is an architecture diagram of the online collaboration method between APC and RTO of the uranium enrichment auxiliary process air conditioning box system provided by the present invention. DETAILED DESCRIPTION

[0063] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0064] The application provides an APC and RTO online cooperation method of an auxiliary process of uranium concentration, which uses APC multivariable cooperation control technology to convert a multivariable control problem between refrigerating capacity, air supply capacity, hall temperature and hall dew point temperature in an air conditioning system into an air conditioning operation condition optimization problem with a constraint adjustment, and the optimal control amount trajectory with the minimum penalty function value is obtained based on the output of a prediction model and the optimization of the penalty function, and the first step of the optimal trajectory is used to issue to an actuator to realize single closed loop control. APC In the next execution period T APC , the optimal trajectory of the minimum value of the penalty function is re-optimized based on the prediction, and the first step of the optimal trajectory is used to issue. In each execution period, the optimal trajectory of the last period is discarded, and the strategy of re-predicting and optimizing according to new feedback is usually called rolling optimization, and the design of rolling ensures that each new feedback information is updated to the optimization problem, so that the temperature and humidity optimization control of the air conditioning system can respond to the current feedback in time, and also can correct in advance based on the prediction of the future, so as to improve the control precision (that is, reduce the temperature fluctuation range).

[0065] The application is innovatively designed on the basis of the APC controller, and the RTO (Real Time Optimization) optimization design is increased on the premise of ensuring the control precision, the energy-saving index is innovatively expressed through the objective function design of the RTO, and finally the output of the RTO optimization participates in the temperature and humidity optimization control of the APC controller, so that the adaptive adjustment and energy-saving optimization control of the auxiliary process system control parameters are realized. The core is that the control target y s (process hall environment temperature and dew point temperature) is optimized in real time through the RTO while the bottom layer control parameters (compressor load and fan frequency) of the auxiliary process system are dynamically optimized through the APC, the output y s of the RTO optimization is the control target of the APC controller, so that the double-layer optimization cooperation of the system APC and RTO and the dynamic adjustment work point are realized to realize energy saving. The output of the RTO optimization is the target value ys of the controlled object of the APC. If the target ys is also rapidly changed, the prediction of the optimization problem to the future is continuously changed, so that the future target is inaccurate when the minimum penalty function trajectory is optimized based on the prediction trajectory, so the execution period T RTO of the RTO optimization is longer, and is usually 10 times of the APC execution period T APC .

[0066] When designing the APC controller, the response characteristics of MV / DV as input to the output CV and the response requirements of the controlled target (such as disturbance recovery time) are expressed based on a multivariable model group table, and the feedforward / feedback mode parameters, speed parameters, disturbance suppression parameters and the like in the APC controller parameters are adjusted to realize accurate control of the closed-loop future controlled variable CV and real-time multivariable optimization control. In the APC implementation of the air conditioning control system of the uranium enrichment cascade hall, the first step of the optimal trajectory of the air valve opening degree and the fan frequency output by the APC at each T APC time point is closed-loop controlled to ensure the stability of the temperature target. At the same time, the control target of CV is optimized by learning the data in the RTO period through RTO, so as to further realize energy saving by dynamically adjusting the working point on the premise of improving the overall temperature and humidity experience of the user.

[0067] As Figure 1 shown, the method of the present application specifically comprises the following steps:

[0068] Step 1, APC algorithm modeling

[0069] An intelligent control model of the air conditioning system is established through the APC algorithm, and the design of the multivariable model predictive controller (APC) is completed. The steps of model modeling are as follows:

[0070] Step 1.1, determining the main related CV / MV / DV variables of the system

[0071] The first step of using the APC algorithm to design the optimization controller needs to determine the main controlled objects and influencing factors of the system, which are usually divided into three categories of variables: the manipulated variable (MV) is the set value of the actuator in the system that can be changed, the controlled variable (CV) and the disturbance variable (DV) are measurable or can be changed (such as the outlet water temperature setting). Table 1 lists the common CV / MV / DV variables of the front end and the end of the central air conditioning system.

[0072] Table 1 list of common CV / MV / DV of central air conditioning system

[0073]

[0074] According to the configuration of different air conditioning system sites, including the sensor, whether the actuator is adjustable or the on-off quantity, the adjustable MV variable, the adjustable or measurable DV variable and the CV variable that need to be controlled are selected.

[0075] The controlled variables, manipulated variables and disturbance variables considered by the APC controller of the cascade hall of Shaanxi Uranium Phase II and Phase III are as follows:

[0076] Air supply temperature T of air conditioning box sa1 ,...,T saNu , return air dew point temperature T dew1 ,...,T dewNu Temperature T of hall temperature and humidity measuring point H1 ,...,T HNy , and hall humidity H H1 ,...,H HNy will be the controlled variable CV, where Nu is the number of air conditioning boxes participating in the optimization control, each air conditioning box measures the supply air temperature and return air dew point temperature, and Ny is the number of temperature and humidity measuring points without faults participating in the control optimization.

[0077] Fan frequency [F1,... F Nu ] of air conditioning box and air conditioning box load rate setting [L1...L Nu ] will be the operation variable MV, where Nu is the number of air conditioning boxes participating in the optimization control.

[0078] Outdoor temperature and humidity disturbance (T out , H out ) will be the main disturbance signal DV.

[0079] Because the number of air conditioning boxes N u and N y are different, each MV and CV is actually a one-to-many relationship.

[0080] Step 1.2, obtain the original model group of the air conditioning system according to model identification experiments

[0081] The CV / MV / DV variables determined in step 1.1 are subjected to corresponding model identification experiments; assuming there are n MVs, m DVs and p CVs, manually set the initial value of each MV to the target value of the air conditioning system to reach the steady state condition and maintain for a certain period of time, and then start the identification experiment. Each time, only change the set value of 1 MV or DV, perform step test, record all CV / MV / DV real-time data and save. The step test of each MV or DV can be continuously performed multiple times with different amplitudes of upward or downward step changes, record the system change data after the change, and then apply the system identification algorithm to identify the model for each MV / DV change p times, thereby obtaining the respective array model of (n+m) MV / DV to p CV. The model identification experiment distribution is shown in Table 2.

[0082] Table 2 Model identification experiment distribution table

[0083] CV1 CV2 ... CVp MV1 Model 1 1 Model 12 Model 1 p MV2 Model 21 Model 22 Model 2 p ... MVn DVn+1 DVn+2 .. DVn+m Model n+m, 1 Model n+m, 2 Model n+m, p

[0084] The time response data is obtained through experiments, and is used for system identification. The model filled in each matrix unit in Table 2 is an input-output transfer function model in the frequency domain space. The entire matrix constitutes a group of original models required for APC control design. The mathematical expression is shown in equation (1).

[0085] y(s) = G(s)u(s) (1)

[0086] where y is (y1, y2, …, yp) corresponding to p CV variables; u is (u1, u2, …, un, un+1, un+m) corresponding to n MVs and m DVs; y(s) and u(s) represent the frequency domain expressions of the input and output variables, and G(s) is a transfer function in the frequency domain.

[0087]

[0088] Step 1.3. Calculate the steady-state gain according to the original model group, and simplify the model of the air conditioning system

[0089] According to the results of the original model group in step 1.2, the steady-state gain of each matrix unit is calculated. Models with no gain or small gain (a commonly used small criterion is that the gain value is less than one-tenth of the maximum gain in the same row, and the criterion can be adjusted according to the requirements in actual application) are removed, thereby obtaining a simplified model group of the air conditioning system (the removed units are filled with 0). The mathematical expression of the multivariable model G(s) can be simplified to Gs(s), and the matrix units with no model or small gain are removed. In the example, the G12 and Gp1 models are modified to 0.

[0090]

[0091] Step 1.4. Convert the continuous frequency model Gs(s) to a discrete space transfer function by a continuous-to-discrete transformation method (such as bilinear transformation)

[0092] The response time of the APC system is selected, the sampling time Tapc of the controller is designed, and the bilinear transformation from continuous to discrete is performed to obtain the discrete space transfer function model G(z) corresponding to the simplified model Gs(s). Equation (2) is the input-output model relationship after discretization.

[0093] y(k) = G(z)u(k) (2)

[0094] Step 2. Design the penalty function and execution period T of the APC optimization APC

[0095] The APC optimization strategy introduces the target of the controlled variable into the penalty function of APC optimization, and takes the outdoor environment temperature and humidity disturbance into the algorithm model. Through real-time optimization control of the input and output of the air conditioning system, the energy-saving operation of the air conditioning system is realized.

[0096] Step 2.1 Designing the APC penalty function and adjustable parameters of the function

[0097] Firstly, the penalty function J(APC) of APC optimization needs to be designed, which is mathematically expressed as follows:

[0098]

[0099] satisfies

[0100] y low ≤y(k+l|k)≤y high l=1,2,…,H cy

[0101] u low ≤u(k+r-1|k)≤u high r=1,2,…,H cu

[0102] Δu(k+i|k)=u(k+i)-u(k+i-1)

[0103] |Δu(k+i-1|k)|≤Δu limit i=1,2,…,H u (5)

[0104] In the formula:

[0105] y low is the lower limit of the output variable (controlled variable), and y high is the upper limit of the output variable (controlled variable);

[0106] u low is the lower limit of the input variable (execution quantity), and u high is the upper limit of the input variable (execution quantity);

[0107] Δu limit is the maximum change rate.

[0108] The symbol (*|k) in formula (4) represents predicting the input / output variable at time k+1, k+2,..., k+x in the kth control period, where * represents the future time starting from the current time; and H u is called the control horizon, that is: the time span of the input sequence ; and H yis called Prediction Horizon. The value of Hy in actual application is mainly designed according to the response characteristics of the system. If the slowest system is expected to reach the stable target value ys after 30 steps, then after 30 steps, each term y(k+i|k)-ys(k+i|k) in JAPC is equal to 0, and there is no need to continue to accumulate the items of 0. Hu in actual application is also corresponding to the specified step, and the output can be kept at the target value. The value of Hu can be equal to or less than the value of Hy. In the case of less than, the input and output time delay is corresponding. The effect of Hu moment can affect the output only after Hu+ time delay, so some actual application scenarios can keep the value of input u unchanged in advance.

[0109] The three inequality constraints in formula (5) are used to cover the range constraints of sensors, actuators and other physical quantities, as well as the response speed constraints of actuators. Hcy represents the future step at which the constraint condition on the output variable takes effect, Hcu represents the future step at which the constraint condition on the input variable takes effect, and Hu is the prediction step of the input variable in formula (4). l, r, i are respectively the time index of k+x moment when the mathematical expression of the constraint condition is expressed. The value of Hcy can be equal to or less than Hy, and the value of Hcu can be equal to or less than Hu.

[0110] The penalty function of J(APC) is composed of the sum of three parts. The first part represents the weighted 2-norm of the deviation of the output y and ys(target) in the future Hy time interval, Q1 is the weight matrix of the output y, the second part represents the weighted 2-norm of the deviation of the input u and us(target) in the future Hu time interval, Q2 is the weight matrix of the input u, and the third part represents the weighted 2-norm of the change of u in the future Hu time interval, Q3 is the weight matrix of the change rate of the input u. The eigenvalues of all weight matrices Q1 / Q2 / Q3 are greater than or equal to 0, so the weighted 2-norm term is also greater than or equal to 0, and the minimum value of J(APC) is 0. According to the size of the temperature control accuracy range of the customer demand of the hall air conditioner, the recovery time after disturbance, the values of Q1, Q2 and Q3 are selected.

[0111] Since J(APC) is designed to use weighted 2-norm, the optimization problem can be solved by quadratic programming (QP, Quadratic Programming) problem. At the same time, the upper and lower limit of the optimization input and output is set respectively y low ,y high , u low , u highThe final optimal solution requires eliminating combinations that exceed the range of y and u. This results in a constrained input / output optimization problem for each cycle of the APC controller. Using a QP solver algorithm, we can quickly find the value of u that satisfies the constraints for each step in the future time period.

[0112] Step 2.2 Design the execution cycle of APC and the corresponding step parameters

[0113] For APC controller, each time T APC At the execution time, it is necessary to solve a future input variable sequence corresponding to the minimum value of J(APC) that meets the constraints. The APC optimization cycle is selected based on the time response characteristics of the air conditioning system model and the user's desired disturbance recovery timeframe. A smaller cycle isn't necessarily better. If sensor readings barely change within a cycle, the optimized output won't change either. If the cycle is too slow, the system might complete the change before the controller has time to execute. Therefore, a cycle of one-tenth of the system's open-loop ramp-up time or less is often chosen. For the Shaanxi Uranium auxiliary control cascade hall air conditioning system, a one-minute cycle was chosen.

[0114] In addition, based on the J(APC) penalty function expressed in the designed quadratic programming, for small and medium-sized systems with less than 50 inputs / 50 outputs, the QP problem can find the optimal solution within 1-3 seconds, ensuring that the optimized updated output can be obtained with a 1-minute execution cycle.

[0115] Step 3: Design RTO optimization penalty function, weight coefficient and RTO period

[0116] The output of APC optimization is based on current feedback and future predictions, stabilizing the controlled variable at the target value while satisfying the optimal input requirements for other response characteristics. The target value of the controlled variable during optimization is a constant ys. RTO optimization, on the other hand, uses data to identify the optimal operating point for the air conditioner, specifically whether ys can be adjusted. Adjusting ys is also expressed through the optimization of an RTO penalty function, which satisfies the optimal combination with the minimum value of a constrained JRTO penalty function.

[0117] Step 3.1 Design of RTO penalty function and weight parameters

[0118] According to the design requirements of the uranium enrichment cascade hall, the spatial distribution of the dry-bulb temperature in the hall is stable, condensation of equipment is avoided, and energy-saving operation is achieved. The RTO optimization penalty function J(RTO) designed in the present invention expresses the weighted sum of the absolute values ​​of multiple temperature and humidity sensors and the target value. Its mathematical expression is as shown in formula (3):

[0119]

[0120] satisfy

[0121]

[0122]

[0123] where J(RTO) is the penalty function of RTO, Ny is the number of hall temperature and humidity sensors participating in optimization, such as 40 temperature and 40 humidity sensors in Shaanxi Uranium Phase 3 hall; i represents the sensor subscript.

[0124] t k，i is the average temperature measurement of the i-th sensor in the k-th call of RTO, which has been processed by singular value removal and low-pass filtering;

[0125] h k，i is the average humidity measurement of the i-th sensor in the k-th call of RTO, which has also been processed by singular value removal and low-pass filtering;

[0126] T tgt is the temperature target, H tgt is the humidity target value, which is also an input item of RTO optimization.

[0127] Through the solution of the linear programming, the optimal air conditioning box control target, i.e. the temperature target T tgt and the humidity target H tgt , can be found.

[0128] q k，i , r k，i are the weight coefficients of temperature deviation T and humidity deviation H, respectively, which are non-negative numbers. The greater the weight value, the greater the influence on the penalty function value. Subscript k is used to distinguish the time of weight coefficient, and the weight value at different times can be adjusted. Subscript i is used to distinguish the weight of different sensor positions on the optimization target.

[0129] T tgt Tmin is the lower bound of the allowed temperature target;

[0130] Tmax is the upper bound of the allowed temperature target;

[0131] H tgt Hmin is the lower bound of the allowed humidity target;

[0132] Hmax is the upper bound of the allowed humidity target.

[0133] The mathematical expression of J(RTO) of the uranium enrichment air conditioning system is the summation of the absolute values of the temperature and humidity deviation values of multiple measuring points, and is a linear mathematical expression. Moreover, the weight coefficients are non-negative numbers, so the minimum penalty function value is 0 in theory. The penalty function optimization problem can be solved by a linear programming method (LP, Linear Programming) to obtain the optimal T tgt and H tgt .

[0134] Step 3.2 Design of the execution period and weight parameters of the RTO penalty function

[0135] According to the optimization output of the RTO, the temperature set value and the enthalpy set value are output, so the RTO optimization period is usually more than 10 times the APC period. At the same time, the closed-loop response time of the air supply fan frequency to the hall temperature and the transition time after the start-stop change of the outdoor compressor of the air conditioning box, the RTO period of the uranium enrichment cascade hall is designed to be 60 times the APC period (i.e. 1 hour).

[0136] In the J(RTO) penalty function, the corresponding time k corresponds to T RTO is updated once. The weight values q and r represent the influence of a certain sensor at a certain time, and the greater the weight value, the higher the control accuracy requirement of the corresponding sensor at the corresponding time. The subscript k is used to distinguish that the weight values at different times can be different. When there is a significant difference in refrigeration effect in different air conditioning seasons, the weight can be adjusted by the operator. The weight of the sensor at different positions can be fine-tuned based on the spatial position, such as increasing the weight of the sensor close to the main process equipment and reducing the weight close to the surrounding walls. At the same time, the weights of temperature and humidity are different, the control accuracy requirement of temperature is high, and the humidity control range is relatively large. When the uranium enrichment air conditioning system is optimized by RTO, the designed q weight value is 3-5 times the r weight value.

[0137] The present application solves the following problems through RTO optimization:

[0138] The dynamic optimization problem of finding the optimal operating point of the air conditioning box is converted into a linear programming problem with constraint adjustment, solving the problem that it is difficult to master the optimal operating point of the air conditioning box under manual control. The dynamic means that the optimal operating point is recalculated every RTO period.

[0139] The weight q k，i , r k，i is a time-varying adjustment weight, supporting dynamic time-varying adjustment, so that the operator can flexibly set the weight value under different seasons or working conditions according to the actual outdoor season or environmental temperature and humidity, thereby changing the influence degree of the dynamic optimization of the working point under different temperature and humidity measuring points.

[0140] The adjustment of dynamic weight can be manually adjusted by operating personnel or automatically switched to different weight combinations when the trigger condition is met through the design of logic and trigger condition, thereby reducing the workload of manual monitoring and adjustment.

[0141] The call period of RTO is set to more than 10 times of the APC optimization call period to ensure that the set target of RTO is fully expressed in the APC optimization objective function.

[0142] Step 4, initialize APC and RTO optimization parameters and start continuous period operation

[0143] After importing the model of step 1 and setting the optimization problem parameters of step 2 and step 3 optimization problem parameters, start the APC and RTO collaborative optimization controller operation.

[0144] Step 4.1 uses the manually set T tgt and H tgt initial value to start APC execution.

[0145] As a necessary parameter of the APC optimization problem, the first time the system starts execution uses the manually set T tgt and H tgt initial value, reads the current time value of the corresponding CV / MV / DV, and completes the first solution of the APC optimization problem. After obtaining the optimal solution that satisfies the constraint condition , only the u(k|k) value is used to issue to the actuator as the set value.

[0146] Step 4.2 reads new feedback values every T APC time, performs a J(APC) optimization problem solution, and uses the first step output value of the optimal solution to issue to the actuator. Step 4.2 embodies the strategy of time rolling optimization.

[0147] Step 4.3 enables J(RTO) optimization solution every T RTO time, and then uses the optimized T tgt and H tgt values to update the ys value of the APC controller. The APC controller always uses the new ys value as the control target until the next RTO optimization period arrives.

[0148] At every T RTO time, J(RTO) optimization and J(APC) optimization are performed at the same time, and J(APC) optimization still uses the previous T tgt and H tgt values.

[0149] Step 4.4 is the first T RTO after T APCAt this moment, use the T updated at the previous moment tgt and H tgt value, repeat step 4.2 to solve the J(APC) optimization problem once, and apply the optimal output of the first step.

[0150] The present invention achieves the minimum deviation between the measured values ​​and targets of multiple controlled objects CV through the coordinated control of RTO and APC. APC Fast cycle and T RTO The coordination of slow cycles, when applied to the air conditioning system of Shaanxi Uranium Phase 3, each T RTO Always keep the optimized control target T tgt With humidity control target H tgt Send to APC real-time optimization subsystem. APC will be based on the fan frequency [F1(k-1)…F Nu (k-1)] and the previous step air-conditioning box load rate setting [L1(k-1)…L Nu (k-1)], and the current temperature measurement value [t1(k)…T Ny (k)] and the current humidity measurement values ​​[h1(k)…h Ny (k)], as well as changes in external disturbances, dynamically output the optimal current fan frequency [F1(k)…F Nu (k)] and the current air-conditioning box load rate setting [L1(k)…L Nu (k)].

[0151] When the fan frequency and air conditioning box load rate set values ​​are sent to the combined air conditioning system in real time, the equipment makes corresponding adjustments, thereby achieving closed-loop control of the air conditioning system.

[0152] Under the special process requirements of the uranium enrichment auxiliary process air conditioning box, this patent makes the following inventions and innovations in the coordinated optimization of RTO and APC:

[0153] (1) Air conditioning box RTO and APC level collaborative optimization. Through the RTO layer, the APC control target of the auxiliary process air conditioning system is dynamically optimized (wherein the optimization includes two levels, linear programming solution based on the RTO penalty function in the RTO optimization period, and quadratic optimization solution based on the APC penalty function in the APC optimization period; the dynamic refers to the optimal result is used only once after each optimization, and the next period is re-optimized to achieve). Through dynamic optimization, the predicted results can be automatically corrected in the presence of model deviation, so that the controller has better robustness. The robustness of the controller refers to the ability of the controller to ensure the closed-loop control effect when the model and the actual system have deviation. Because new data is updated and the optimization problem is solved again in each calculation period, the robustness of the controller is improved, the precision of the controlled variable is better, the system closed-loop commissioning rate is improved, and finally the air conditioning system is realized. Unmanned intelligent control while greatly reducing the energy consumption of the air conditioning system.

[0154] (2) Through real-time optimization control of RTO, the target values of the hall temperature and humidity are optimized and solved in each T RTO period, and are issued to the APC controller of the air conditioning box. The working point of the air conditioning box is collaboratively optimized;

[0155] (3) Through APC optimization control, optimization based on feedback and model prediction is realized, which has better feedforward effect and disturbance suppression, and improves the control precision of the system. For example, the temperature precision of the air conditioning system of Shaanxi Uranium Phase 3 is controlled from + / - 1.5 degrees to + / - 1 degree.

[0156] The APC and RTO online collaborative method provided by the application can realize intelligent closed-loop regulation of the air conditioning box and effectively reduce the daily operation amount after being put into use.

[0157] According to the actual production situation, the closed-loop rate is generally counted once every natural month, and if the system running set condition is changed before and after the change, the closed-loop rate is counted before and after the change, respectively. The closed-loop rate calculation formula is as follows:

[0158] Closed-loop rate = (APC control time put into use) / (cumulative time of APC control put into use + cumulative time of manual intervention control) * 100%.

[0159] The operation amount is measured and calculated according to the cumulative manual intervention time in the statistical time period;

[0160] Operation amount reduction rate = (1- (manual intervention time in the statistical period after using the optimization control) / cumulative manual intervention time in the statistical period before using the optimization control)) * 100%.

[0161] After the manual intervention time is proposed in the statistical period, the measurement variable sample (n) of the hall temperature and the hall humidity is put into automatic operation control, the absolute value of the difference between each sample and the target of the temperature / humidity is counted, the sample number n (1) and n (1.5) of the absolute value of the difference.

[0162] After the method optimization technology of the present application is used, n (1) / n*100% is greater than 95.45%, and it can be considered that the positive and negative 1 degree is 2 sigma, and 95.45% probability satisfies the deviation less than 1 degree.

[0163] Before the method optimization technology of the present application is used, n (1.5) / n*100% is greater than 95.45%, but n (1) / n*100% is less than 95%, so the precision improvement is verified.

[0164] The control performance improvement is reflected in the precision improvement, that is, (1.5-1) / 1.5*100%=33%.

[0165] The present application is described in detail above in combination with the drawings and examples, but the present application is not limited to the above examples, and various changes can be made within the knowledge possessed by those skilled in the art without departing from the purpose of the present application. The contents not described in detail in the present application can adopt the existing technology.

Claims

1. A method for online collaboration between APC and RTO in a uranium enrichment auxiliary process air conditioning system, characterized in that: The method comprises the following steps: Step 1: APC algorithm modeling; Step 2: Design the penalty function and execution cycle T for APC optimization APC ; Step 3: Design the RTO optimization penalty function, weight coefficient, and RTO period; Step 4: Initialize the APC and RTO optimization parameters and start continuous cycle operation; The step 2 includes: Step 2.1 Design the penalty function for APC optimization and the adjustable parameters of the function; Step 2.2 Design the execution cycle of APC and the corresponding step size parameters; The mathematical expression of the penalty function J(APC) optimized by APC is: satisfy Δu(k+i|k)=u(k+i)-u(k+i-1) Where: y low is the lower bound of the output variable, y high is the upper bound of the output variable; u low is the lower bound of the input variable, u high is the upper bound of the input variable; Δu limit is the maximum rate of change.

2. The online collaboration method of APC and RTO of uranium enrichment auxiliary process air conditioning box system according to claim 1 is characterized in that: The step 1 comprises: Step 1.1, determine the system's manipulated variable MV, controlled variable CV and disturbance variable DV; Step 1.2: Obtain the original model group of the air-conditioning system according to the model identification experiment; Step 1.3: Simplify the air conditioning system model based on the statistical steady-state gain of the original model group; Step 1.4: Convert the continuous frequency model into a transfer function in discrete space through the continuous-to-discrete transformation method.

3. The online collaboration method of APC and RTO of uranium enrichment auxiliary process air conditioning box system according to claim 2 is characterized in that: The step 1.2 is specifically as follows: conducting a corresponding model identification experiment on the CV / MV / DV variables determined in step 1.1; changing only the set value of one MV or DV at a time, performing a step test, recording and saving the real-time data of all CV / MV / DVs; performing the step test on each MV or DV by continuously performing multiple upward or downward step changes of different amplitudes, recording the change data of the system after the change, applying the system identification algorithm, performing p model identifications on each group of MV / DV changes, and obtaining respective array models of (n+m) MV / DV pairs of p CVs.

4. The online collaboration method of APC and RTO of a uranium enrichment auxiliary process air conditioning box system according to claim 2, characterized in that: The step 1.3 is specifically as follows: according to the original model group results of step 1.2, the steady-state gain of each matrix unit is counted, and the models without models or with small gains are eliminated to obtain a simplified model group of the air-conditioning system.

5. The online collaboration method of APC and RTO of uranium enrichment auxiliary process air conditioning box system according to claim 1 is characterized in that: The step 3 includes: Step 3.1 Design of RTO penalty function and weight parameters; Step 3.2 Design of the execution cycle and weight parameters of the RTO penalty function.

6. The online collaboration method for APC and RTO of a uranium enrichment auxiliary process air conditioning box system according to claim 5, characterized in that: The mathematical expression of the RTO penalty function in step 3.1 is: satisfy Where J(RTO) is the penalty function of RTO, Ny is the number of hall temperature and humidity measurement points participating in the optimization, and i represents the sensor subscript; t k，i is the average temperature measurement value of the i-th measurement point when the RTO is called for the kth time; h k，i is the average humidity measurement value of the i-th measurement point when the RTO is called for the kth time; T tgt is the temperature target, H tgt is the humidity target value; q k，i , r k，i are the weight coefficients of temperature deviation T and humidity deviation H respectively; H tgt is the lower bound of the allowed temperature target; is the upper limit of the allowed temperature target; H tgt is the lower bound of the allowed humidity target; The upper limit of the allowed humidity target.

7. The online collaboration method of APC and RTO of uranium enrichment auxiliary process air conditioning box system according to claim 1 is characterized in that: The step 4 comprises: Step 4.1 Using manually set T tgt and H tgt The initial value is given to the APC controller to start APC execution; Step 4.2 Each T APC At this moment, the new feedback value is read, the J(APC) optimization problem is solved once, and the first step output value with the best solution is sent to the actuator; Step 4.3 At each T RTO At this moment, enable the optimization solution of J(RTO), and then use the optimized output T tgt and H tgt The APC controller uses the new ys value to update the ys value. Before the next RTO optimization cycle comes, the APC controller always uses the new ys value as the control target. Step 4.4 In T RTO The first T after the moment APC At this moment, use the T updated at the previous moment tgt and H tgt value, repeat step 4.2 to solve the J(APC) optimization problem once, and apply the optimal output of the first step.

Images