Model predictive control system of heat-conducting oil furnace

Abstract

The invention relates to a model prediction control system of a heat conduction oil furnace. The model prediction control system comprises a model prediction module, a rolling optimization module and a feedback correction module, the model prediction module predicts the change condition of the control variable of the control system within a period P in the future under the control action of the moment k by utilizing a first-order lag transfer function model according to the parameters of the current operation variable, the control variable and the disturbance variable, and outputs the change condition to the rolling optimization module; the rolling optimization module calculates an optimal control increment of an operation variable in a future period of time P under the control action of the moment k by utilizing an iterative optimization mode; and the feedback correction module compares the prediction result at the (k + 1) moment with the actual output of the control system at the same moment, solves the deviation between the prediction result and the actual output, and corrects the model prediction result at the (k + 1) moment by using the deviation. According to the method, the change conditions of the heat conduction oil outlet temperature and the flue gas residual oxygen can be effectively predicted, and the stability of the heat conduction oil outlet temperature and the flue gas residual oxygen is ensured.

Description

Technical Field

[0001] The invention belongs to the technical field of thermal oil furnace design and relates to a model prediction control system of a thermal oil furnace. Background Art

[0002] A thermal oil boiler is a boiler used for thermal oil circulation heating and heat transfer. The medium is heated by burning oil, gas, coal and other fuels through a burner, and then a high-temperature oil pump is used for liquid phase circulation to transport the heated medium to the heat-using equipment. The outlet temperature of the thermal oil boiler and the residual oxygen content of the flue gas are the core control parameters. In the chemical production process, the traditional automatic control method is double cross cascade PID feedback control. This control method can achieve point-to-point control and can better control the outlet temperature of the thermal oil boiler to be stable, meet the user's heat demand, and is suitable for simple linear systems and heat-stable scenarios.

[0003] However, when the outlet temperature or flow rate of the heat transfer oil is disturbed and fluctuates, the PID control loop of the outlet temperature of the heat transfer oil furnace and the residual oxygen content in the flue gas has a control lag for the fluctuation and cannot be adjusted in time, which can easily cause the fuel and air volume regulating valves to oscillate or even fully open or close. Therefore, it is necessary to develop a model predictive control system to predict the changes in the outlet temperature of the heat transfer oil and the residual oxygen content in the flue gas, and take control measures in advance to ensure the stability of the outlet temperature of the heat transfer oil and the residual oxygen content in the flue gas. Summary of the invention

[0004] The technical problem solved by the invention is to overcome the deficiencies of the prior art and provide a model prediction control system for a thermal oil furnace.

[0005] The solution of the present invention is:

[0006] A model prediction control system for a thermal oil furnace, comprising a model prediction module, a rolling optimization module, and a feedback correction module;

[0007] Model prediction module: Based on the current operating variables, control variables and disturbance variable parameters, the first-order lag transfer function model is used to predict the changes in the control system control variables within a period of time P in the future under the control action at time k, and output them to the rolling optimization module; the operating variables are gas flow and air volume, the control variables are heat transfer oil outlet temperature and flue gas residual oxygen content; the disturbance variables include heat transfer oil flow, heat transfer oil inlet temperature, and exhaust gas temperature;

[0008] Rolling optimization module: It uses iterative optimization to calculate the optimal control increment of the operating variable within a period of time P in the future under the control action at time k;

[0009] Feedback correction module: compares the predicted result at time k+1 with the actual output of the control system at the same time, finds the deviation between the two, and uses the deviation to correct the model prediction result at time k+1.

[0010] Preferably, the first-order lag transfer function model includes:

[0011] The first-order lag transfer function of the current gas flow and air volume to the thermal oil outlet temperature;

[0012] The first-order lag transfer function of the current gas flow and air volume to the residual oxygen content of the flue gas;

[0013] The first-order lag transfer function of the heat transfer oil flow rate to the heat transfer oil outlet temperature;

[0014] The first-order lag transfer function of the heat transfer oil inlet temperature to the heat transfer oil outlet temperature;

[0015] The first-order lag transfer function of exhaust gas temperature to thermal oil outlet temperature;

[0016] The first-order lag transfer function is in the form of

[0017]

[0018] Where K, T, and τ are parameters, K represents the transfer function gain, and T represents the time constant;

[0019] τ represents the delay parameter, which is the time from the action of the control variable to the beginning of the change of the manipulated variable.

[0020] Preferably, it also includes a model identification module for obtaining parameters of the first-order lag transfer function model through online identification or historical data.

[0021] Preferably, the model identification module obtains the parameters of the first-order lag transfer function model through an online identification method, and the method is as follows:

[0022] Perform step excitation test on the first-order lag transfer function model;

[0023] Modify the gas flow, air volume, heat transfer oil flow, heat transfer oil inlet temperature, and exhaust temperature respectively to obtain the corresponding heat transfer oil outlet temperature change curve and flue gas residual oxygen content change curve when the gas flow, air volume, heat transfer oil flow, heat transfer oil inlet temperature, and exhaust temperature change;

[0024] The parameters of the first-order lag transfer function model are calculated according to the above change curve.

[0025] Preferably, the model identification module obtains the parameters of the first-order lag transfer function model through historical data in the following way:

[0026] Collect the thermal oil outlet temperature, fuel flow, air volume, flue gas residual oxygen content, fuel regulating valve opening, air volume frequency setting opening, exhaust temperature, thermal oil inlet temperature, and thermal oil flow of the thermal oil furnace M days before, and obtain the corresponding thermal oil outlet temperature change curve and flue gas residual oxygen content change curve when the gas flow, air volume, thermal oil flow, thermal oil inlet temperature, and exhaust temperature change according to the above data;

[0027] The parameters of the first-order lag transfer function model are calculated according to the above-mentioned change trend curve.

[0028] Preferably, the predicted output value of the model prediction module is satisfy:

[0029]

[0030] Right now

[0031]

[0032] in, is the predicted output value of the control variable at the next P moments under the control action at the k moment, is the initial predicted value of the control variable at the next P moments under the control action at moment k, Δu M (k) is the control increment of the manipulated variable at the next M moments under the control action at moment k; A is the model information matrix, a 1 、a 2 , ...a P They are the steady-state values ​​of the response of the first sampling moment, the second sampling moment, and ... the Pth sampling moment of the first-order lag transfer function model respectively.

[0033] Preferably, the rolling optimization module calculates the optimal control increment of the operating variable within a future period P under the control action at time k by an iterative optimization method, and the method is as follows:

[0034] The optimization performance index J(k) at time k is constructed by using the penalty tracking error and the adjustment amplitude:

[0035]

[0036] Where q i and r j Represent the error weight coefficient and control weight coefficient respectively; APC_CV1(k+i) represents the expected value of the control variable at time k+i, represents the predicted output value of the controlled variable at time k+i under the control action at time k, and Δu is the control increment of the manipulated variable at time (k+j-1) under the control action at time k;

[0037] Find the minimum value of J(k) when dJ(k) / dΔuM When (k) = 0, we can obtain Δu M The optimal value of (k).

[0038] Preferably, the feedback correction module implements the formula as follows:

[0039]

[0040] In the formula is the predicted process value of the control variable after weighted correction of the prediction error at the next N moments under the control action at the k moment, is the initial predicted value of the control variable at the next N moments under the control action at moment k, is the initial predicted value of the control variable at the next N moments under the control action at the k+1 moment, represents the correction coefficient vector, represents the transfer matrix; e(k+1) represents the error between the true value and the predicted value at time k+1.

[0041] The beneficial effects of the present invention compared with the prior art are:

[0042] (1) More stable: Since the fuel amount and air volume calculated by the model prediction are used as the limiting parameters of the temperature PID loop SV adjustment, the fuel amount, air volume and air-fuel ratio are always adjusted within a reasonable amplitude range, so the dynamic process of the new temperature control system is more stable.

[0043] (2) Faster: When interference variables such as the inlet temperature and flow rate of the thermal oil furnace fluctuate, the model can be used to predict future temperature changes before the outlet temperature of the thermal oil changes. Control measures can be taken in advance to adjust the gas and air volumes to further stabilize the outlet temperature of the thermal oil. This is faster than traditional control methods.

[0044] (3) More optimized: Since the prediction weight matrix and control weight matrix are introduced in the rolling optimization process, the prediction output and control output can be optimized when calculating the optimal control law, ensuring the stability of the gas and air volumes, and stabilizing the heat transfer oil outlet temperature and flue gas residual oxygen content with the most optimized control. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a working schematic diagram of the present invention;

[0046] Figure 2 This is a flow chart of the rolling optimization module of the present invention;

[0047] Figure 3 This is a logic diagram of the model prediction module of the present invention;

[0048] Figure 4 This is a logic diagram of the feedback correction module of the present invention;

[0049] Figure 5 This is a diagram of the configuration screen of the host computer according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The present invention will be further described below in conjunction with the accompanying drawings.

[0051] The model prediction control system in the present invention comprises a model identification module, a model prediction module, a rolling optimization module and a feedback correction module.

[0052] Model identification module: used to obtain the parameters of the first-order lag transfer function model through online identification or historical data.

[0053] Model prediction module: Based on the current operating variables, control variables and disturbance variable parameters, the first-order lag transfer function model is used to predict the changes in the control system control variables within a period of time P in the future under the control action at time k, and output them to the rolling optimization module; the operating variables are the gas flow rate and air volume, the control variables are the heat transfer oil outlet temperature and the flue gas residual oxygen content; the disturbance variables include the heat transfer oil flow rate, the heat transfer oil inlet temperature, and the exhaust gas temperature.

[0054] Rolling optimization module: It uses iterative optimization to calculate the optimal control increment of the operating variable within a future period P under the control action at time k.

[0055] Feedback correction module: compares the predicted result at time k+1 with the actual output of the control system at the same time, finds the deviation between the two, and uses the deviation to correct the model prediction result at time k+1.

[0056] The first-order lag transfer function model includes:

[0057] The first-order lag transfer function of the current gas flow and air volume to the thermal oil outlet temperature;

[0058] The first-order lag transfer function of the current gas flow and air volume to the residual oxygen content of the flue gas;

[0059] The first-order lag transfer function of the heat transfer oil flow rate to the heat transfer oil outlet temperature;

[0060] The first-order lag transfer function of the heat transfer oil inlet temperature to the heat transfer oil outlet temperature;

[0061] First-order lag transfer function of exhaust gas temperature to thermal oil outlet temperature.

[0062] The first-order lag transfer function is of the form

[0063]

[0064] Where K, T, and τ are parameters, K represents the transfer function gain, and T represents the time constant;

[0065] The model identification module obtains the parameters of the first-order lag transfer function model through an online identification method as follows:

[0066] Perform step excitation test on the first-order lag transfer function model;

[0067] Modify the gas flow, air volume, heat transfer oil flow, heat transfer oil inlet temperature, and exhaust temperature respectively to obtain the corresponding heat transfer oil outlet temperature change curve and flue gas residual oxygen content change curve when the gas flow, air volume, heat transfer oil flow, heat transfer oil inlet temperature, and exhaust temperature change;

[0068] The parameters of the first-order lag transfer function model are calculated according to the above change curve.

[0069] The model identification module can also obtain the parameters of the first-order lag transfer function model through historical data as follows:

[0070] Collect the thermal oil outlet temperature, fuel flow, air volume, flue gas residual oxygen content, fuel regulating valve opening, air volume frequency setting opening, exhaust temperature, thermal oil inlet temperature, and thermal oil flow of the thermal oil furnace M days before, and obtain the corresponding thermal oil outlet temperature change curve and flue gas residual oxygen content change curve when the gas flow, air volume, thermal oil flow, thermal oil inlet temperature, and exhaust temperature change according to the above data;

[0071] The parameters of the first-order lag transfer function model are calculated according to the above-mentioned change trend curve.

[0072] The method of the present invention is as follows:

[0073] (1) The model identification module obtains the parameters of the first-order lag transfer function model;

[0074] (2) The model prediction module uses the first-order lag transfer function model to predict the change of the control variable of the control system in the future period P under the control action at time k based on the current operating variables, control variables and disturbance variable parameters and the initial prediction of the control variable at time k;

[0075] The rolling optimization module uses iterative optimization to calculate the optimal control increment of the operating variable within a future period P under the control action at time k;

[0076] The feedback correction module compares the predicted result at time k+1 with the actual output of the control system at the same time, finds the deviation between the two, and uses the deviation to calculate the initial predicted value of the control variable at time k+1;

[0077] (3) Repeat step (2) at each subsequent moment to achieve real-time prediction.

[0078] Example:

[0079] Figure 1 It is a working schematic diagram of the present invention. Figure 2 This is a flow chart of the rolling optimization module of the present invention. Figure 3 This is the logic diagram of the model prediction module of the present invention. Figure 4 This is the logic diagram of the feedback correction module of the present invention.

[0080] 1 Hardware control system design

[0081] The hardware of the model predictive control system adopts the mode of controller and host computer. The controller adopts programmable logic controller PLC, PLC uses Siemens S7-1500 series, PLC includes high-performance CPU and communication interface, configures the model predictive control algorithm program block, and compiles and downloads the configured program to the PLC controller. The host computer uses DELL industrial computer + Siemens WINCC V7.5 RT-2048 software. The host computer and controller are connected through a network cable, and the communication protocol is industrial Ethernet.

[0082] 2 Module design

[0083] (1) Step excitation test

[0084] The stimulus test is implemented by combining historical curve query and artificial step test.

[0085] (2) Model Identification Module

[0086] Check the trend curves of the thermal oil outlet temperature, fuel flow, air volume, flue gas residual oxygen content, fuel regulating valve opening, air volume frequency setting opening, exhaust temperature, heat medium furnace inlet temperature, and heat medium flow rate of the thermal oil furnace in the first 20 days, find out the historical values ​​of the above parameters under several typical operation cycles, and manually calculate K, T, τ required for the first-order lag transfer function model:

[0087]

[0088] In formula 1, K represents the transfer function gain. T represents the time constant, which is the time when the controlled variable reaches 63.2% of the steady state. τ represents the delay parameter, which is the time required for the controlled variable to start changing when the control variable is in action.

[0089] The transfer function model of the operating variable (gas flow and air volume) to the control variable (heat transfer oil outlet temperature), K, T, τ values ​​are 0.02, 150, 120. The transfer function model of the operating variable (gas flow and air volume) to the control variable (flue gas residual oxygen content), K, T, τ values ​​are 0.05, 130, 100.

[0090] The transfer function model of the disturbance variable 1 (heat transfer oil flow) to the control variable (heat transfer oil outlet temperature), K, T, τ values ​​are -0.082, 120, 60. The transfer function model of the disturbance variable 2 (heat transfer oil inlet temperature) to the control variable (heat transfer oil outlet temperature), K, T, τ values ​​are 1.4, 120, 60. The transfer function model of the disturbance variable 3 (exhaust gas temperature) to the control variable (heat transfer oil outlet temperature), K, T, τ values ​​are -0.2, 120, 60.

[0091] (3) Model prediction module

[0092] The model prediction module predicts the future output based on the step response of the currently known transfer function and assumes future inputs. Therefore, the model prediction value is the sum of the free term (zero input response) and the forced term (zero state response).

[0093]

[0094] in, is the predicted output value of the control variable at the next P moments under the control action at the k moment, is the initial predicted value of the control variable at the next P moments under the control action at moment k, Δu M (k) is the control increment of the manipulated variable at the next M moments under the control action at moment k; A is the model information matrix, a 1 、a 2 , ...a P They are the steady-state response values ​​of the first sampling moment, the second sampling moment, and ... the Pth sampling moment of the first-order lag transfer function model respectively.

[0095] In formula 2, Predict the output value for P dimension. is the initial prediction value of P dimension. A is the model information matrix (PxM dimension), Δu M (k) is the control increment of the M-dimensional manipulated variable, obtained through online optimization.

[0096] (4) Rolling Optimization Module

[0097] Control increment Δu of the manipulated variable M (k) Perform iterative optimization and use the penalty tracking error and adjustment amplitude to construct the optimization performance index:

[0098]

[0099] In formula 3, q i and r j Represent the error weight coefficient and control weight coefficient respectively. APC_CV1(k+i) represents the expected value of the control variable at time k+i. represents the predicted output value of the controlled variable at time k+i under the control action at time k, and Δu is the control increment of the manipulated variable at time (k+j-1) under the control action at time k;

[0100] Find the minimum value of J(k) when dJ(k) / dΔu M When (k) = 0, we can obtain Δu M The optimal value of (k).

[0101] (5) Feedback Correction Module

[0102] The error of the previous prediction is used to compensate the predicted value of the future output, making the predicted value closer to the true value. The main principle formula is as follows:

[0103]

[0104] In formula 4 is the predicted process value of the control variable after weighted correction of the prediction error at the next N moments under the control action at the k moment, is the initial predicted value of the control variable at the next N moments under the control action at moment k, is the initial predicted value of the control variable at the next N moments under the control action at the k+1 moment, represents the correction coefficient vector, represents the transfer matrix; e(k+1) represents the error between the true value and the predicted value at time K+1.

[0105] This embodiment is based on the TIA Portal SCL programming language, and designs and writes a model prediction control algorithm, including a model prediction module, a rolling optimization control module, and a feedback correction module. First, the parameters in each program block are initialized, and the current operating variables, control variables, and disturbance variable parameters of the control system are input into the model prediction program block, and the changes in the controlled variables of the control system within a period of time P in the future are calculated. Because the system is always in the process of change, each prediction only predicts the changes in the controlled variables of the system under the control action at the current k moment. At the k+1 moment, the control action will not be the optimal control law, so the k moment needs to be re-optimized and calculated to calculate the optimal control law at the k+1 moment. The results calculated by the model prediction are input into the rolling optimization program block, and the optimal control law at the k+1 moment is calculated to act on the control system. The current control output belongs to open-loop prediction, and there will be some deviations between the predicted results and the actual output of the control system. Therefore, the predicted result y(k+1|k) is compared with the output y(k+1) of the control system at the same time, and the deviation e(k) between the two is obtained, and e(k) is applied to the correction of the model prediction results at the next moment to ensure the accuracy of the model prediction results. Repeat the above three steps at each subsequent moment to achieve model predictive control.

[0106] In this embodiment,

[0107] (1) A function block is written for the "rolling optimization module", and the necessary port definitions are as follows:

[0108]

[0109] (2) A function block is written for the "model prediction module", and the necessary port definitions are as follows:

[0110]

[0111] (3) A function block is written for the “Feedback Correction Module” and the necessary port definitions are as follows:

[0112]

[0113]

[0114] (4) Display the human-computer interaction parameters required by the model prediction system on the panel of the host computer, such as Figure 5 As shown, the user first confirms whether there is an alarm of the measuring instrument. If the indicator light behind "Instrument Alarm" shows green, it means that the instrument functions normally. Then press the button behind "Main Switch". When the button shows "On", it means that the model prediction function has entered the commissioning state. Finally, pre-enter the set values ​​of the thermal oil boiler outlet temperature and the residual oxygen content of the flue gas, and select the "on" and "off" states of the operating variables and interference variables according to the actual working conditions. The model predictive control system can be put into operation and the model predictive control system is put into operation.

[0115] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A model predictive control system for a thermal oil furnace, characterized in that: Including model prediction module, rolling optimization module, feedback correction module; Model prediction module: Based on the current operating variables, control variables and disturbance variable parameters, the first-order lag transfer function model is used to predict the changes in the control system control variables within a period of time P in the future under the control action at time k, and output them to the rolling optimization module; the operating variables are gas flow and air volume, the control variables are heat transfer oil outlet temperature and flue gas residual oxygen content; the disturbance variables include heat transfer oil flow, heat transfer oil inlet temperature, and exhaust gas temperature; Rolling optimization module: It uses iterative optimization to calculate the optimal control increment of the operating variable within a period of time P in the future under the control action at time k; Feedback correction module: compares the predicted result at time k+1 with the actual output of the control system at the same time, finds the deviation between the two, and uses the deviation to correct the model prediction result at time k+1.

2. The model predictive control system for a thermal oil furnace according to claim 1, characterized in that: The first-order lag transfer function model includes: The first-order lag transfer function of the current gas flow and air volume to the thermal oil outlet temperature; The first-order lag transfer function of the current gas flow and air volume to the residual oxygen content of the flue gas; The first-order lag transfer function of the heat transfer oil flow rate to the heat transfer oil outlet temperature; The first-order lag transfer function of the heat transfer oil inlet temperature to the heat transfer oil outlet temperature; The first-order lag transfer function of exhaust gas temperature to thermal oil outlet temperature; The first-order lag transfer function is in the form of Where K, T, and τ are parameters, K represents the transfer function gain, and T represents the time constant; τ represents the delay parameter, which is the time from the action of the control variable to the beginning of the change of the manipulated variable.

3. The model predictive control system for a thermal oil furnace according to claim 1, characterized in that: The invention also comprises a model identification module for obtaining the parameters of the first-order lag transfer function model through online identification or historical data.

4. The model predictive control system for a thermal oil furnace according to claim 3, characterized in that: The model identification module obtains the parameters of the first-order lag transfer function model through an online identification method as follows: Perform step excitation test on the first-order lag transfer function model; Modify the gas flow, air volume, heat transfer oil flow, heat transfer oil inlet temperature, and exhaust temperature respectively to obtain the corresponding heat transfer oil outlet temperature change curve and flue gas residual oxygen content change curve when the gas flow, air volume, heat transfer oil flow, heat transfer oil inlet temperature, and exhaust temperature change; The parameters of the first-order lag transfer function model are calculated according to the above change curve.

5. The model predictive control system for a thermal oil furnace according to claim 3, characterized in that: The model identification module obtains the parameters of the first-order lag transfer function model through historical data as follows: Collect the thermal oil outlet temperature, fuel flow, air volume, flue gas residual oxygen content, fuel regulating valve opening, air volume frequency setting opening, exhaust temperature, thermal oil inlet temperature, and thermal oil flow of the thermal oil furnace M days before, and obtain the corresponding thermal oil outlet temperature change curve and flue gas residual oxygen content change curve when the gas flow, air volume, thermal oil flow, thermal oil inlet temperature, and exhaust temperature change according to the above data; The parameters of the first-order lag transfer function model are calculated according to the above-mentioned change trend curve.

6. The model predictive control system for a thermal oil furnace according to claim 1, characterized in that: The predicted output value of the model prediction module satisfy: Right now in, is the predicted output value of the control variable at the next P moments under the control action at the k moment, is the initial predicted value of the control variable at the next P moments under the control action at moment k, Δu M (k) is the control increment of the manipulated variable at the next M moments under the control action at moment k; A is the model information matrix, a1, a2, ... a P They are the steady-state values ​​of the response of the first sampling moment, the second sampling moment, and ... the Pth sampling moment of the first-order lag transfer function model respectively.

7. The model predictive control system for a thermal oil furnace according to claim 1, characterized in that: The rolling optimization module uses an iterative optimization method to calculate the optimal control increment of the operating variable within a future period P under the control action at time k, and the method is as follows: The optimization performance index J(k) at time k is constructed by using the penalty tracking error and the adjustment amplitude: Where q i and r j Represent the error weight coefficient and control weight coefficient respectively; APC_CV1(k+i) represents the expected value of the control variable at time k+i, represents the predicted output value of the controlled variable at time k+i under the control action at time k, and Δu is the control increment of the manipulated variable at time (k+j-1) under the control action at time k; Find the minimum value of J(k) when dJ(k) / dΔu M When (k) = 0, we can obtain Δu M The optimal value of (k).

8. The model predictive control system for a thermal oil furnace according to claim 1, characterized in that: The feedback correction module is implemented as follows: In the formula is the predicted process value of the control variable after weighted correction of the prediction error at the next N moments under the control action at the k moment, is the initial predicted value of the control variable at the next N moments under the control action at moment k, is the initial predicted value of the control variable at the next N moments under the control action at the k+1 moment, represents the correction coefficient vector, represents the transfer matrix; e(k+1) represents the error between the true value and the predicted value at time k+1.

Images