Advanced coordination control method for reheat steam temperature system of coal-fired unit based on data driving

Abstract

The invention discloses an advanced coordination control method for a reheat steam temperature system of a coal-fired unit based on data driving, and the method is based on a data driving model, combines and fuses a multi-input multi-output implicit generalized predictive control algorithm, and achieves the coordination control of the temperature of reheat steam through temperature reduction water spraying and flue gas baffle adjustment. The method comprises the steps that firstly, steam temperature set values of a reheat steam inlet and a reheat steam outlet are calculated through PID and a fitting formula, and then initial control instruction increments of the opening degree of a desuperheater control valve and the opening degree of a smoke baffle are calculated through a multiple-output-multiple-output implicit generalized predictive control algorithm; 2, a feedforward instruction is obtained through the boiler master control instruction; 3, obtaining a feedforward instruction of the predicted value of the reheat steam temperature; the desuperheater control valve opening instruction is obtained through incremental calculation and amplitude limiting of an initial control instruction, and the flue gas baffle opening instruction is obtained through summation calculation and amplitude limiting of the initial control instruction increment and a feedforward instruction. The method is of great significance in improving the rapid load changing capacity of the coal-fired unit.

Description

Technical Field

[0001] The present invention belongs to the technical field of thermal power system control optimization, and in particular relates to an advanced coordinated control method for a reheat steam temperature system of a coal-fired unit based on data drive. Background Art

[0002] With the transformation and upgrading of my country's energy system, coal-fired units are required to undertake more peak-shaving and frequency-regulating tasks to accommodate more renewable energy sources such as solar and wind power to ensure the safe and stable operation of the power grid. Therefore, coal-fired units need to more frequently operate in rapid load changes and wide load ranges.

[0003] The quality of reheat steam temperature control has become a significant factor limiting the ability of coal-fired power generation thermal power systems to further improve their variable load capabilities. Low reheat steam temperatures reduce the thermal cycle efficiency of coal-fired units, impacting their safe operation. Excessively high reheater temperatures reduce the safety factor of metal tube strength, lowering the thermal cycle efficiency of coal-fired units. Therefore, achieving more stable reheat steam temperature control during rapid load changes is an urgent issue. Summary of the Invention

[0004] The purpose of the present invention is to address the deficiencies of the above-mentioned prior art and provide a data-driven advanced coordinated control method for the reheat steam temperature system of a coal-fired unit to optimize the reheat steam temperature control system of the coal-fired unit, and to achieve coordinated control of the reheat steam temperature by water spray cooling and flue gas dampers through advanced control methods, thereby achieving better reheat steam temperature control.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] An advanced coordinated control method for the reheat steam temperature system of a coal-fired unit based on data drive is proposed. The acquisition of the flue gas damper opening instruction of the reheat steam temperature system of the coal-fired unit includes three parts: the initial control instruction increment of the flue gas damper calculated based on the implicit generalized predictive control algorithm of the multi-input and multi-output system, the feedforward instruction generated by the boiler master control instruction, and the feedforward instruction generated by the predicted value of the steam temperature at the last stage reheater inlet. The desuperheater valve opening instruction of the reheat steam temperature system of the coal-fired unit is obtained by cascade control, which includes two parts: the first part is the desuperheater outlet temperature set value generated by the last stage reheater outlet temperature deviation, and the second part is the initial control instruction increment of the desuperheater valve opening calculated based on the implicit generalized predictive control algorithm of the multi-input and multi-output system.

[0007] The flue gas damper opening and the desuperheater valve opening are inputs, and the final reheater inlet and outlet steam temperatures are outputs, forming a two-input and two-output control system in the following form:

[0008]

[0009] Where: (z -1 ) is the backshift operator; A1(z -1 )、A2(z -1 ), B 11 (z -1 ), B 12 (z -1 ), B 21 (z -1 ) and B 22 (z -1 ) are polynomials containing the parameters to be identified; y1(t) is the final reheater inlet steam temperature output at the current time t; y2(t) is the final reheater outlet steam temperature output at the current time t; u1(t) is the flue gas damper opening input at the current time t; u2(t) is the desuperheater valve opening input at the current time t; Δ = 1-z -1 is the difference operator; ξ1(k),ξ2(k) are both uncorrelated random sequences, representing the influence of a type of random noise, which can be ignored;

[0010] The first step is to calculate the initial control command increment, which is mainly obtained from the final reheater inlet steam temperature, final reheater outlet steam temperature, flue gas damper opening at the previous moment, desuperheater valve opening at the previous moment, and the final reheater inlet and outlet steam temperature set values ​​generated by the boiler master control command operation:

[0011] (1) Calculate the set value of the outlet steam temperature of the final reheater by fitting the quadratic calculation formula as follows:

[0012] F1(x)=a1*DOB 2 +b1*DOB+c1

[0013] Where: F1(x) is the set value of the outlet steam temperature of the final reheater; a1 is the constant coefficient of the quadratic term; DOB is the boiler master control instruction; b1 is the constant coefficient of the linear term; c1 is the constant term coefficient;

[0014] (2) Calculate the set value of the steam temperature at the inlet of the final reheater. This step is the calculation of the outer loop of the desuperheater cascade control. The set value of the steam temperature at the inlet of the final reheater is obtained by PID calculation based on the deviation between the steam temperature at the outlet of the final reheater and the set value of the steam temperature at the outlet of the final reheater F1(x) calculated in the previous step:

[0015] e rh =F1(x)-T rh

[0016] Δu rh (t) = K p *(e rh (t)-erh (t-1))+K i *e rh (t)

[0017] +K d *(e rh (t)-2*e rh (t-1)+e rh (t-2))

[0018] u rh (t)=Δu rh (t)+u rh (t-1)

[0019] Where: e rh is the outlet steam temperature deviation of the last stage reheater; T rh is the outlet steam temperature of the last stage reheater, °C; Δu rh (t) is the current moment of the last stage reheater inlet steam temperature set value instruction increment; K p is the proportional coefficient; e rh (t) is the steam temperature deviation at the outlet of the last stage reheater at the current moment; e rh (t-1) is the outlet steam temperature deviation of the last stage reheater at the previous moment; K i is a number; K d is the differential coefficient; e rh (t-2) is the outlet steam temperature deviation of the last stage reheater at the previous moment; u rh (t) is the current set value of the final reheater inlet steam temperature; u rh (t-1) is the set value of the steam temperature at the inlet of the last stage reheater at the previous moment;

[0020] (3) Parameter identification of a two-input and two-output control system, which is decomposed into two subsystems:

[0021] A1(z -1 )y1(k)=B 11 (z -1 )u1(k)+B 12 (z -1 )u2(k)+ξ1(k) / Δ

[0022] A2(z -1 )y2(k)=B 21 (z -1 )u1(k)+B 22 (z -1 )u2(k)+ξ2(k) / Δ

[0023] Use recursive least squares to perform parameter identification on the two subsystems separately:

[0024]

[0025] Where: is the vector of parameters to be identified at the current moment, g n-1 ,g n-2 ,…,g0 is the identified parameter, f(t) is the initial value predicted at the current moment; It is the vector of parameters to be identified at the previous moment. The initial value is usually K(t) is a constant matrix, usually K(t)=10 6 I, I are the unit matrices; y(t) is the actual SCR outlet NOx concentration at the current moment; n is the maximum prediction length; X(tn) is the vector consisting of the ammonia injection amount command increments at the previous n moments; P(t) and P(t-1) are the constant matrices at the current moment and the previous moment, respectively; λ1 is the forgetting factor;

[0026] (4) Calculate the initial control command increment. The objective function of the two-input and two-output control system is as follows:

[0027]

[0028] Where: J is the objective function; m is the control length; t is the current time; j is the sequence number;

[0029] is the output at time t+j, y1(t+j) and y2(t+j) correspond to the steam temperature output values ​​at the inlet and outlet of the last stage reheater respectively; is the reference trajectory at time t+j, ω1(t+j) and ω2(t+j) are the reference trajectories of the steam temperature at the inlet and outlet of the last stage reheater at time t+j, a is the softening factor, y r1 ,y r2 are the steam temperature set points at the inlet and outlet of the last stage reheater respectively; λ(j) is the control weighting coefficient at time t+j; is the control increment at time t+j-1, u1(t+j-1) and u2(t+j-1) are the control increments of the flue gas damper opening and the desuperheater valve opening at time t+j-1 respectively;

[0030] On this basis, the control law is obtained, and the initial control instruction increment is obtained through the control law operation:

[0031]

[0032] Where: ΔU1=[Δu1(t),Δu1(t+1),…,Δu1(t+n-1)] Tis the control increment input vector of the smoke damper opening, where Δu(t), Δu(t+1), and Δu(t+n-1) correspond to the control increment inputs of the smoke damper opening at time t, t+1, and t+n-1, respectively; ΔU2=[Δu2(t),Δu2(t+1),…,Δu2(t+n-1)] T is the control increment input vector of the desuperheater valve opening, where Δu(t), Δu(t+1) and Δu(t+n-1) correspond to the control increment input of the desuperheater valve opening at time t, t+1 and t+n-1 respectively; G 11 , G 12 , G 21 and G 22 The matrix is ​​the lower triangular matrix composed of the identification parameters in step (3); I is the identity matrix; W1=[ω1(t+1),ω1(t+2),…,ω1(t+n)] T is the reference trajectory matrix of the final reheater inlet and outlet steam temperatures, where ω1(t+1), ω1(t+2), and ω1(t+n) correspond to the reference trajectories of the final reheater inlet steam temperature at time t+1, t+2, and t+n, respectively; W2 = [ω2(t+1), ω2(t+2), …, ω2(t+n)] T is the reference trajectory matrix of the final reheater outlet steam temperature, where ω1(t+1), ω1(t+2) and ω1(t+n) correspond to the reference trajectories of the final reheater outlet steam temperature at time t+1, t+2 and t+n respectively; λ is the control weighting coefficient; f1 is the prediction initial value matrix of the final reheater inlet steam temperature; f2 is the prediction initial value matrix of the final reheater outlet steam temperature. The prediction initial value can be obtained by using the output prediction value of the previous moment and feedback correction to obtain the prediction initial value of the current moment. Both prediction initial value matrices can be calculated according to the following formula:

[0033]

[0034] Where: e(t) is the prediction error at the current time t; y(t) is the NOx concentration output at the SCR outlet at the current time t; represents the output prediction value of the previous time t-1 for the current time t, and so on; f(t), f(t+1) and f(t+n-1) are the initial prediction values ​​of the corresponding time; h1, h2 and h n is the error correction coefficient;

[0035] The output prediction value is calculated by the following formula:

[0036]

[0037] Where: is a vector matrix composed of the predicted values ​​of the steam temperature at the last stage reheater inlet at different times, where and The output prediction values ​​of the last stage reheater inlet steam temperature at time t+1, t+2 and t+n respectively correspond to; is a vector matrix composed of the predicted values ​​of the outlet steam temperature of the last stage reheater at different times, where and The output prediction values ​​of the outlet steam temperature of the last stage reheater at time t+1, t+2 and t+n respectively correspond to the output prediction values ​​of the outlet steam temperature of the last stage reheater at time t+1, t+2 and t+n;

[0038] The vectors ΔU1 and ΔU2 are obtained by calculation, and their first terms Δu1(t) and Δu2(t) are taken respectively, which are the initial control instruction increment ΔCMD of the corresponding smoke damper instruction. fg (t) The initial control command increment ΔCMD of the desuperheater valve opening command jw (t);

[0039] The second step is to calculate the feedforward command generated by the boiler master control command. The calculation formula is as follows:

[0040] F2(x)=k2*DOB

[0041] Where: F2(x) is the feedforward instruction generated by the boiler master control instruction; k2 is the calculation coefficient;

[0042] The third step is to calculate the feedforward instruction generated by the predicted value of the final reheater inlet steam temperature. The predicted value of the final reheater inlet steam temperature is obtained by the long short-term memory network prediction model with the fusion attention mechanism. First, the time series data of the parameters affecting the final reheater inlet steam temperature are obtained, and the erroneous values ​​are eliminated. The time series data with non-equal time intervals and default values ​​are resampled using the piecewise cubic spline interpolation method to obtain equal time interval data for modeling. Secondly, the data features are screened using the principal component analysis method, and the features with a high impact on the final reheater inlet steam temperature are retained as model inputs. Then, the delay of the input parameters is estimated using the mutual information method, and the parameters are aligned to avoid the reduction in prediction accuracy caused by the delay time. Finally, the processed input and output are used to establish the long short-term memory network prediction model with the fusion attention mechanism for predicting the final reheater inlet steam temperature.

[0043] The feedforward instruction generated by the predicted value of the final reheater inlet steam temperature is obtained by the following formula:

[0044] F3(x)=k3*T rh-in

[0045] Where: F3(x) is the feedforward instruction generated by the predicted value of the final reheater inlet steam temperature; k3 is the calculation coefficient; Trh-in is the predicted value of the steam temperature at the last stage reheater inlet.

[0046] The fourth step is to calculate the desuperheater valve opening instruction and the flue gas damper opening instruction;

[0047] (1) Desuperheater valve opening instruction, calculated as follows:

[0048] CMD jw (t) = CMD jw (t-1)+ΔCMD jw (t)

[0049] Where: CMD jw (t) is the desuperheater valve opening instruction at the current moment; CMD jw (t-1) is the desuperheater valve opening instruction at the previous moment; ΔCMD jw (t) is the initial control increment of the desuperheater valve opening instruction at the current moment, which is calculated in the first step;

[0050] (2) Smoke damper opening instruction, the calculation formula is as follows:

[0051] CMD fg (t) = CMD fg (t-1)+ΔCMD fg (t)+F2(x)+F3(x)

[0052] Where: CMD fg (t) is the smoke damper opening instruction at the current moment; CMD fg (t-1) is the smoke damper opening instruction at the previous moment; ΔCMD fg (t) is the initial control increment of the smoke damper opening instruction at the current moment, which is calculated in the first step.

[0053] Preferably, when establishing a long short-term memory network prediction model that integrates an attention mechanism for predicting the final reheater inlet steam temperature, the parameter characteristics involved in the water spray desuperheating process that directly affects the reheat steam temperature of the coal-fired unit are considered, including the desuperheating water volume and the steam temperature at the water spray desuperheating inlet. The primary air volume, secondary air volume, and coal feed rate involved in the boiler combustion process that affect the temperature are also considered. The parameter characteristics of the high-pressure cylinder exhaust pressure and steam extraction volume on the turbine side are also considered. Based on these direct or indirect parameter characteristics, relevant operating data is collected, and this data is required to be time series data with equal time intervals. A training dataset is constructed based on this data. This solution systematically selects the desuperheating water parameter characteristics, boiler combustion parameter characteristics, and turbine-side parameter characteristics that are directly related to the final reheater inlet steam temperature, constructing a dataset with clear physical meaning, and ultimately achieving high-precision prediction of the final reheater inlet steam temperature under complex operating conditions.

[0054] Preferably, the calculated desuperheater valve opening instruction satisfies the following relationship:

[0055] CMD jw-min <CMD jw <CMD jw-max

[0056] Where: CMD jw CMD is the desuperheater valve opening instruction; jw-min The lower limit of the desuperheater valve opening; CMD jw-max The upper limit of the desuperheater valve opening.

[0057] Preferably, the calculated flue gas damper opening instruction satisfies the following relationship:

[0058] CMD fg-min <CMD fg <CMD fg-max

[0059] Where: CMD fg It is the flue gas damper opening instruction; CMD fg-min The lower limit of the flue gas damper opening instruction; CMD fg-max It is the upper limit of the flue gas damper opening instruction.

[0060] Compared with the prior art, the present invention has the following advantages:

[0061] (1) The method of the present invention is aimed at the reheat steam temperature control system with strong nonlinearity and uncertainty. It realizes the coordinated control of reheat steam temperature by water spray cooling and flue gas damper through multi-input multi-output advanced control algorithm.

[0062] (2) Based on the prediction model of the steam temperature at the final reheater inlet, the present invention constructs a feedforward instruction based on data drive and optimizes the reheat steam temperature control;

[0063] (3) The present invention integrates advanced control methods and data-driven models to effectively address the large delay and large inertia problems of the reheat steam temperature of coal-fired units and optimize the control effect of the reheat steam temperature. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is the logic diagram of the optimized reheat steam temperature coordinated control strategy of the present invention.

[0065] Figure 2 This is a flowchart of the data-driven prediction model mentioned in the present invention.

[0066] Figure 3 (a) and (b) are comparison results of the mean value and standard deviation of the outlet steam temperature of the last stage reheater under the control of the method proposed by the present invention and the traditional method, respectively. DETAILED DESCRIPTION

[0067] The present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0068] The present invention proposes a data-driven advanced coordinated control method for the reheat steam temperature system of a coal-fired unit. The specific implementation method is as follows:

[0069] like Figure 1 As shown in the figure, the acquisition of the flue gas damper opening instruction of the reheat steam temperature system of the coal-fired unit includes three parts, namely, the initial control instruction increment of the flue gas damper calculated based on the implicit generalized predictive control algorithm of the multi-input and multi-output system, the feedforward instruction generated by the boiler master control instruction, and the feedforward instruction generated by the predicted value of the steam temperature at the last stage reheater inlet; the desuperheater regulating valve opening instruction of the reheat steam temperature system of the coal-fired unit is obtained by cascade control, which includes two parts. The first part is to generate the desuperheater outlet temperature set value through the last stage reheater outlet temperature deviation, and the second part is the initial control instruction increment of the desuperheater regulating valve opening calculated based on the implicit generalized predictive control algorithm (IGPC) of the multi-input and multi-output system.

[0070] The flue gas damper opening and the desuperheater valve opening are inputs, and the final reheater inlet and outlet steam temperatures are outputs, forming a two-input and two-output control system in the following form:

[0071]

[0072] Where: (z -1 ) is the backshift operator; A1(z -1 )、A2(z -1 ), B 11 (z -1 ), B 12 (z -1 ), B 21 (z -1 ) and B 22 (z -1 ) are polynomials containing the parameters to be identified; y1(t) is the final reheater inlet steam temperature output at the current time t; y2(t) is the final reheater outlet steam temperature output at the current time t; u1(t) is the flue gas damper opening input at the current time t; u2(t) is the desuperheater valve opening input at the current time t; Δ = 1-z -1 is the difference operator; ξ1(k),ξ2(k) are both uncorrelated random sequences, representing the influence of a type of random noise, which can be ignored;

[0073] S01, calculate the initial control command increment, which is mainly obtained by the final reheater inlet steam temperature, final reheater outlet steam temperature, flue gas damper opening at the previous moment, desuperheater valve opening at the previous moment, and the final reheater inlet and outlet steam temperature set values ​​generated by the boiler master control command operation:

[0074] (1) Calculate the set value of the outlet steam temperature of the final reheater by fitting the quadratic calculation formula as follows:

[0075] F1(x)=a1*DOB 2 +b1*DOB+c1

[0076] Where: F1(x) is the set value of the outlet steam temperature of the final reheater; a1 is the constant coefficient of the quadratic term; DOB is the boiler master control instruction; b1 is the constant coefficient of the linear term; c1 is the constant term coefficient;

[0077] (2) Calculate the set value of the steam temperature at the inlet of the final reheater. This step is the calculation of the outer loop of the desuperheater cascade control. The set value of the steam temperature at the inlet of the final reheater is obtained by PID calculation based on the deviation between the steam temperature at the outlet of the final reheater and the set value of the steam temperature at the outlet of the final reheater F1(x) calculated in the previous step:

[0078] e rh =T rh-set -T rh

[0079] Δu rh (t) = K p *(e rh (t)-e rh (t-1))+K i *e rh (t)

[0080] +K d *(e rh (t)-2*e rh (t-1)+e rh (t-2))

[0081] u rh (t)=Δu rh (t)+u rh (t-1)

[0082] Where: e rh is the outlet steam temperature deviation of the last stage reheater; T rh is the outlet steam temperature of the last stage reheater, °C; Δu rh (t) is the current moment of the last stage reheater inlet steam temperature set value instruction increment; K p is the proportional coefficient; e rh(t) is the steam temperature deviation at the outlet of the last stage reheater at the current moment; e rh (t-1) is the outlet steam temperature deviation of the last stage reheater at the previous moment; K i is a number; K d is the differential coefficient; e rh (t-2) is the outlet steam temperature deviation of the last stage reheater at the previous moment; u rh (t) is the current set value of the final reheater inlet steam temperature; u rh (t-1) is the set value of the steam temperature at the inlet of the last stage reheater at the previous moment;

[0083] (3) Parameter identification of the two-input and two-output control system model, which is decomposed into two subsystems:

[0084] A1(z -1 )y1(k)=B 11 (z -1 )u1(k)+B 12 (z -1 )u2(k)+ξ1(k) / Δ

[0085] A2(z -1 )y2(k)=B 21 (z -1 )u1(k)+B 22 (z -1 )u2(k)+ξ2(k) / Δ

[0086] Recursive least squares (RLS) is used to identify the parameters of the two subsystems respectively:

[0087]

[0088] Where: is the vector of parameters to be identified at the current moment, g n-1 ,g n-2 ,…,g0 is the identified parameter, f(t) is the initial value predicted at the current moment; It is the vector of parameters to be identified at the previous moment. The initial value is usually K(t) is a constant matrix, usually K(t)=10 6 I, I are the unit matrices; y(t) is the actual SCR outlet NOx concentration at the current moment; n is the maximum prediction length; X(tn) is the vector consisting of the ammonia injection amount command increments at the previous n moments; P(t) and P(t-1) are the constant matrices at the current moment and the previous moment, respectively; λ1 is the forgetting factor;

[0089] (4) Calculate the initial control command increment. The objective function of the two-input and two-output control system is as follows:

[0090]

[0091] Where: J is the objective function; m is the control length; t is the current time; j is the sequence number; is the output at time t+j, y1(t+j) and y2(t+j) correspond to the steam temperature output values ​​at the inlet and outlet of the last stage reheater respectively; is the reference trajectory at time t+j, ω1(t+j) and ω2(t+j) are the reference trajectories of the steam temperature at the inlet and outlet of the last stage reheater at time t+j, a is the softening factor, y r1 ,y r2 are the steam temperature set points at the inlet and outlet of the last stage reheater respectively; λ(j) is the control weighting coefficient at time t+j; is the control increment at time t+j-1, u1(t+j-1) and u2(t+j-1) are the control increments of the flue gas damper opening and the desuperheater valve opening at time t+j-1 respectively;

[0092] On this basis, the control law is obtained, and the initial control instruction increment is obtained through the control law operation:

[0093]

[0094] Where: ΔU1=[Δu1(t),Δu1(t+1),…,Δu1(t+n-1)] T is the control increment input vector of the smoke damper opening, where Δu(t), Δu(t+1), and Δu(t+n-1) correspond to the control increment inputs of the smoke damper opening at time t, t+1, and t+n-1, respectively; ΔU2=[Δu2(t),Δu2(t+1),…,Δu2(t+n-1)] T is the control increment input vector of the desuperheater valve opening, where Δu(t), Δu(t+1) and Δu(t+n-1) correspond to the control increment input of the desuperheater valve opening at time t, t+1 and t+n-1 respectively; G 11 , G 12 , G 21 and G 22 The matrix is ​​the lower triangular matrix composed of the identification parameters in step (3); I is the identity matrix; W1=[ω1(t+1),ω1(t+2),…,ω1(t+n)] T is the reference trajectory matrix of the inlet and outlet steam temperatures of the last stage reheater, where ω1(t+1), ω1(t+2) and ω1(t+n) correspond to the reference trajectories of the inlet steam temperature of the last stage reheater at times t+1, t+2 and t+n, respectively;

[0095] W2=[ω2(t+1),ω2(t+2),…,ω2(t+n)] T is the reference trajectory matrix of the final reheater outlet steam temperature, where ω1(t+1), ω1(t+2) and ω1(t+n) correspond to the reference trajectories of the final reheater outlet steam temperature at time t+1, t+2 and t+n respectively; λ is the control weighting coefficient; f1 is the prediction initial value matrix of the final reheater inlet steam temperature; f2 is the prediction initial value matrix of the final reheater outlet steam temperature. The prediction initial value can be obtained by using the output prediction value of the previous moment and feedback correction to obtain the prediction initial value of the current moment. Both prediction initial value matrices can be calculated according to the following formula:

[0096]

[0097] Where: e(t) is the prediction error at the current time t; y(t) is the NOx concentration output at the SCR outlet at the current time t; represents the output prediction value of the previous time t-1 for the current time t, and so on; f(t), f(t+1) and f(t+n-1) are the initial prediction values ​​of the corresponding time; h1, h2 and h n is the error correction coefficient;

[0098] The output prediction value is calculated by the following formula:

[0099]

[0100] Where: is a vector matrix composed of the predicted values ​​of the steam temperature at the last stage reheater inlet at different times, where and The output prediction values ​​of the last stage reheater inlet steam temperature at time t+1, t+2 and t+n respectively correspond to; is a vector matrix composed of the predicted values ​​of the outlet steam temperature of the last stage reheater at different times, where

[0101] and The output prediction values ​​of the outlet steam temperature of the last stage reheater at time t+1, t+2 and t+n respectively correspond to the output prediction values ​​of the outlet steam temperature of the last stage reheater at time t+1, t+2 and t+n;

[0102] The vectors ΔU1 and ΔU2 are obtained by calculation, and their first terms Δu1(t) and Δu2(t) are taken respectively, which are the initial control instruction increment ΔCMD of the corresponding smoke damper instruction. fg (t) The initial control command increment ΔCMD of the desuperheater valve opening command jw (t);

[0103] S02, calculate the feedforward instruction generated by the boiler master control instruction, the calculation formula is as follows:

[0104] F2(x)=k2*DOB

[0105] Where: F2(x) is the feedforward instruction generated by the boiler master control instruction; k2 is the calculation coefficient;

[0106] S03, calculate the feedforward instruction generated by the predicted value of the steam temperature at the final stage reheater inlet, the predicted value of the steam temperature at the final stage reheater inlet is obtained by the long short-term memory network prediction model integrated with the attention mechanism; Figure 2 As shown in the figure, first, the time series data of the parameters affecting the steam temperature at the last stage reheater inlet are obtained, and the erroneous values ​​are eliminated. The time series data with non-equal time intervals and default values ​​are resampled by piecewise cubic spline interpolation method to obtain equal time interval data for modeling. Secondly, the data features are screened by principal component analysis method, and the features with high influence on the steam temperature at the last stage reheater inlet are retained as model input. Then, the delay of the input parameters is estimated by mutual information method, and the parameters are aligned to avoid the reduction of prediction accuracy caused by the delay time. Finally, the processed input and output are used to establish a long short-term memory network prediction model with fusion attention mechanism for the prediction of the steam temperature at the last stage reheater inlet.

[0107] The feedforward instruction generated by the predicted value of the final reheater inlet steam temperature is obtained by the following formula:

[0108] F3(x)=k3*T rh-in

[0109] Where: F3(x) is the feedforward instruction generated by the predicted value of the final reheater inlet steam temperature; k3 is the calculation coefficient; T rh-in is the predicted value of the steam temperature at the last stage reheater inlet.

[0110] S04, calculating the desuperheater valve opening instruction and the flue gas damper opening instruction.

[0111] (1) Desuperheater valve opening instruction, calculated as follows:

[0112] CMD jw (t) = CMD jw (t-1)+ΔCMD jw (t)

[0113] Where: CMD jw (t) is the desuperheater valve opening instruction at the current moment; CMD jw (t-1) is the desuperheater valve opening instruction at the previous moment; ΔCMD jw (t) is the initial control increment of the desuperheater valve opening instruction at the current moment, which is calculated in the first step;

[0114] (2) Smoke damper opening instruction, the calculation formula is as follows:

[0115] CMD fg (t) = CMD fg (t-1)+ΔCMD fg (t)+F2(x)+F3(x)

[0116] Where: CMD fg (t) is the smoke damper opening instruction at the current moment; CMD fg (t-1) is the smoke damper opening instruction at the previous moment; ΔCMD fg (t) is the initial control increment of the smoke damper opening instruction at the current moment, which is calculated in the first step.

[0117] The present invention provides a detailed instruction calculation acquisition method, which calculates each instruction value in sequence according to the control logic order, and collaboratively considers the adjustment of the flue gas damper and the adjustment of the reheat steam temperature by the desuperheater, and combines the data-driven model to predict the final reheater inlet steam temperature as a feedforward. Then, the multi-input and multi-output generalized predictive control algorithm is used to realize the calculation of the control increment, which can capture complex nonlinear and coupling relationships, and finally obtain the flue gas damper opening instruction and the desuperheater valve opening instruction. Thanks to the advantages of the advanced control algorithm in processing the reheat steam temperature control system with large inertia and large delay, as well as the corresponding feedforward, the control effect of the reheat steam temperature is optimized. Figure 3 (a) and (b) are comparisons of the statistical results of the average value and standard deviation of the outlet steam temperature of the last stage reheater under the control of the method proposed in the present invention and the traditional PID method during the load change process. It can be seen that compared with the traditional method, the average value of the method proposed in the present invention is closer to the set value of the outlet steam temperature of the last stage reheater, and the standard deviation under the control of the method proposed in the present invention is smaller, and the discrete degree of the corresponding data is smaller, indicating that the fluctuation degree of the outlet steam temperature of the last stage reheater under the method of the present invention is smaller, and the overall control of the reheat steam temperature is more stable, that is, the strong target tracking feature of the advanced control algorithm is utilized, which can effectively suppress the temperature fluctuation of the reheat steam temperature during the frequent load change of the coal-fired unit, and stably track the set target value, thereby ensuring the stability of the reheat steam temperature, which is of great significance to improving the operating flexibility of the coal-fired unit.

Claims

1. A data-driven advanced coordinated control method for the reheat steam temperature system of a coal-fired unit, characterized by: The acquisition of the flue gas damper opening instruction of the reheat steam temperature system of a coal-fired unit includes three parts: the initial control instruction increment of the flue gas damper calculated based on the implicit generalized predictive control algorithm of the multi-input multi-output system, the feedforward instruction generated by the boiler master control instruction, and the feedforward instruction generated by the predicted value of the steam temperature at the last stage reheater inlet; the desuperheater regulating valve opening instruction of the reheat steam temperature system of a coal-fired unit is obtained by cascade control, which includes two parts: the first part is the desuperheater outlet temperature set value generated by the last stage reheater outlet temperature deviation, and the second part is the initial control instruction increment of the desuperheater regulating valve opening calculated based on the implicit generalized predictive control algorithm of the multi-input multi-output system; The flue gas damper opening and the desuperheater valve opening are inputs, and the final reheater inlet and outlet steam temperatures are outputs, forming a two-input and two-output control system in the following form: Where: (z -1 ) is the backshift operator; A1(z -1 )、A2(z -1 ), B 11 (z -1 ), B 12 (z -1 ), B 21 (z -1 ) and B 22 (z -1 ) are polynomials containing the parameters to be identified; y1(t) is the final reheater inlet steam temperature output at the current time t; y2(t) is the final reheater outlet steam temperature output at the current time t; u1(t) is the flue gas damper opening input at the current time t; u2(t) is the desuperheater valve opening input at the current time t; Δ = 1-z -1 is the difference operator; ξ1(k),ξ2(k) are both uncorrelated random sequences, representing the influence of a type of random noise, which can be ignored; The first step is to calculate the initial control command increment, which is mainly obtained from the final reheater inlet steam temperature, final reheater outlet steam temperature, flue gas damper opening at the previous moment, desuperheater valve opening at the previous moment, and the final reheater inlet and outlet steam temperature set values ​​generated by the boiler master control command operation: (1) Calculate the set value of the outlet steam temperature of the final reheater by fitting the quadratic calculation formula as follows: F1(x)=a1*DOB 2 +b1*DOB+c1 Where: F1(x) is the set value of the outlet steam temperature of the final reheater; a1 is the constant coefficient of the quadratic term; DOB is the boiler master control instruction; b1 is the constant coefficient of the linear term; c1 is the constant term coefficient; (2) Calculate the set value of the steam temperature at the inlet of the final reheater. This step is the calculation of the outer loop of the desuperheater cascade control. The set value of the steam temperature at the inlet of the final reheater is obtained by PID calculation based on the deviation between the steam temperature at the outlet of the final reheater and the set value of the steam temperature at the outlet of the final reheater F1(x) calculated in the previous step: e rh =F1(x)-T rh Δu rh (t)=K p *(e rh (these rh (t-1))+K i *e rh (t) +K d *(e rh (t)-2*e rh (t-1)+e rh (t-2)) u rh (t)=Δu rh (t)+u rh (t-1) Where: e rh is the outlet steam temperature deviation of the last stage reheater; T rh is the outlet steam temperature of the last stage reheater, °C; Δu rh (t) is the current moment of the last stage reheater inlet steam temperature set value instruction increment; K p is the proportional coefficient; e rh (t) is the steam temperature deviation at the outlet of the last stage reheater at the current moment; e rh (t-1) is the outlet steam temperature deviation of the last stage reheater at the previous moment; K i is a number; K d is the differential coefficient; e rh (t-2) is the temperature deviation of the steam outlet of the last stage reheater at the previous moment; is the current set value of the final reheater inlet steam temperature; u rh (t-1) is the set value of the steam temperature at the inlet of the last stage reheater at the previous moment; (3) Parameter identification of a two-input and two-output control system, which is decomposed into two subsystems: A1(with -1 )y1(k)=B 11 (With -1 )u1(k)+B 12 (With -1 )u2(k)+ξ1(k) / Δ A2(z -1 )y2(k)=B 21 (z -1 )u1(k)+B 22 ( z -1 )u2(k)+ξ2(k) / Δ Use recursive least squares to perform parameter identification on the two subsystems separately: Where: is the vector of parameters to be identified at the current moment, g n-1 ,g n-2 ,…,g0 is the identified parameter, f(t) is the initial value predicted at the current moment; It is the vector of parameters to be identified at the previous moment. The initial value is usually K(t) is a constant matrix, usually K(t)=10 6 I, I are the unit matrices; y(t) is the actual SCR outlet NOx concentration at the current moment; n is the maximum prediction length; X(tn) is the vector consisting of the ammonia injection amount command increments at the previous n moments; P(t) and P(t-1) are the constant matrices at the current moment and the previous moment, respectively; λ1 is the forgetting factor; (4) Calculate the initial control command increment. The objective function of the two-input and two-output control system is as follows: Where: J is the objective function; m is the control length; t is the current time; j is the sequence number; is the output at time t+j, y1(t+j) and y2(t+j) correspond to the steam temperature output values ​​at the inlet and outlet of the last stage reheater respectively; is the reference trajectory at time t+j, ω1(t+j) and ω2(t+j) are the reference trajectories of the steam temperature at the inlet and outlet of the last stage reheater at time t+j, a is the softening factor, y r1 ,y r2 are the steam temperature set points at the inlet and outlet of the last stage reheater respectively; λ(j) is the control weighting coefficient at time t+j; is the control increment at time t+j-1, u1(t+j-1) and u2(t+j-1) are the control increments of the flue gas damper opening and the desuperheater valve opening at time t+j-1 respectively; On this basis, the control law is obtained, and the initial control instruction increment is obtained through the control law operation: Where: ΔU1=[Δu1(t),Δu1(t+1),…,Δu1(t+n-1)] T is the control increment input vector of the smoke damper opening, where Δu(t), Δu(t+1), and Δu(t+n-1) correspond to the control increment inputs of the smoke damper opening at time t, t+1, and t+n-1, respectively; ΔU2=[Δu2(t),Δu2(t+1),…,Δu2(t+n-1)] T is the control increment input vector of the desuperheater valve opening, where Δu(t), Δu(t+1) and Δu(t+n-1) correspond to the control increment input of the desuperheater valve opening at time t, t+1 and t+n-1 respectively; G 11 , G 12 , G 21 and G 22 The matrix is ​​the lower triangular matrix composed of the identification parameters in step (3); I is the identity matrix; W1=[ω1(t+1),ω1(t+2),…,ω1(t+n)] T is the reference trajectory matrix of the final reheater inlet and outlet steam temperatures, where ω1(t+1), ω1(t+2), and ω1(t+n) correspond to the reference trajectories of the final reheater inlet steam temperature at time t+1, t+2, and t+n, respectively; W2 = [ω2(t+1), ω2(t+2), …, ω2(t+n)] T is the reference trajectory matrix of the final reheater outlet steam temperature, where ω1(t+1), ω1(t+2) and ω1(t+n) correspond to the reference trajectories of the final reheater outlet steam temperature at time t+1, t+2 and t+n respectively; λ is the control weighting coefficient; f1 is the prediction initial value matrix of the final reheater inlet steam temperature; f2 is the prediction initial value matrix of the final reheater outlet steam temperature. The prediction initial value can be obtained by using the output prediction value of the previous moment and feedback correction to obtain the prediction initial value of the current moment. Both prediction initial value matrices can be calculated according to the following formula: Where: e(t) is the prediction error at the current time t; y(t) is the NOx concentration output at the SCR outlet at the current time t; represents the output prediction value of the previous time t-1 for the current time t, and so on; f(t), f(t+1) and f(t+n-1) are the initial prediction values ​​of the corresponding time; h1, h2 and h n is the error correction coefficient; The output prediction value is calculated by the following formula: Where: is a vector matrix composed of the predicted values ​​of the steam temperature at the last stage reheater inlet at different times, where and The output prediction values ​​of the last stage reheater inlet steam temperature at time t+1, t+2 and t+n respectively correspond to; is a vector matrix composed of the predicted values ​​of the outlet steam temperature of the last stage reheater at different times, where and The output prediction values ​​of the outlet steam temperature of the last stage reheater at time t+1, t+2 and t+n respectively correspond to the output prediction values ​​of the outlet steam temperature of the last stage reheater at time t+1, t+2 and t+n; The vectors ΔU1 and ΔU2 are obtained by calculation, and their first terms Δu1(t) and Δu2(t) are taken respectively, which are the initial control instruction increment ΔCMD of the corresponding smoke damper instruction. fg (t) The initial control command increment ΔCMD of the desuperheater valve opening command jw (t); The second step is to calculate the feedforward command generated by the boiler master control command. The calculation formula is as follows: F2(x)=k2*DOB Where: F2(x) is the feedforward instruction generated by the boiler master control instruction; k2 is the calculation coefficient; The third step is to calculate the feedforward instruction generated by the predicted value of the final reheater inlet steam temperature. The predicted value of the final reheater inlet steam temperature is obtained by the long short-term memory network prediction model with the fusion attention mechanism. First, the time series data of the parameters affecting the final reheater inlet steam temperature are obtained, and the erroneous values ​​are eliminated. The time series data with non-equal time intervals and default values ​​are resampled using the piecewise cubic spline interpolation method to obtain equal time interval data for modeling. Secondly, the data features are screened using the principal component analysis method, and the features with a high impact on the final reheater inlet steam temperature are retained as model inputs. Then, the delay of the input parameters is estimated using the mutual information method, and the parameters are aligned to avoid the reduction in prediction accuracy caused by the delay time. Finally, the processed input and output are used to establish the long short-term memory network prediction model with the fusion attention mechanism for predicting the final reheater inlet steam temperature. The feedforward instruction generated by the predicted value of the final reheater inlet steam temperature is obtained by the following formula: F3(x)=k3*T rh-in Where: F3(x) is the feedforward instruction generated by the predicted value of the final reheater inlet steam temperature; k3 is the calculation coefficient; T rh-in is the predicted value of the steam temperature at the last stage reheater inlet. The fourth step is to calculate the desuperheater valve opening instruction and the flue gas damper opening instruction; (1) Desuperheater valve opening instruction, calculated as follows: CMD jw (t)=CMD jw (t-1)+ΔCMD jw (t) Where: CMD jw (t) is the desuperheater valve opening instruction at the current moment; CMD jw (t-1) is the desuperheater valve opening instruction at the previous moment; ΔCMD jw (t) is the initial control increment of the desuperheater valve opening instruction at the current moment, which is calculated in the first step; (2) Smoke damper opening instruction, the calculation formula is as follows: CMD fg (t)=CMD fg (t-1)+ΔCMD fg (t)+F2(x)+F3(x) Where: CMD fg (t) is the smoke damper opening instruction at the current moment; CMD fg (t-1) is the smoke damper opening instruction at the previous moment; ΔCMD fg (t) is the initial control increment of the smoke damper opening instruction at the current moment, which is calculated in the first step.

2. The data-driven advanced coordinated control method for the reheat steam temperature system of a coal-fired unit according to claim 1 is characterized in that: When establishing a long short-term memory network prediction model with an integrated attention mechanism for predicting the final reheater inlet steam temperature, the parameter characteristics involved in the water spray cooling process that directly affects the reheat steam temperature of the coal-fired unit are considered, including the cooling water volume and the steam temperature at the water spray cooling inlet. The primary air volume, secondary air volume, and coal feed rate involved in the boiler combustion process that affect the temperature are considered. The exhaust pressure of the high-pressure cylinder on the turbine side and the extraction volume parameter characteristics are considered. Based on these direct or indirect parameter characteristics, relevant operating data are collected, and these data are required to be time series data with equal time intervals. A training dataset is constructed based on these data.

3. The data-driven advanced coordinated control method for the reheat steam temperature system of a coal-fired unit according to claim 1 is characterized in that: The calculated desuperheater valve opening command satisfies the following relationship: CMD jw-min <CMD jw <CMD jw-max Where: CMD jw CMD is the desuperheater valve opening instruction; jw-min The lower limit of the desuperheater valve opening; CMD jw-max The upper limit of the desuperheater valve opening.

4. The data-driven advanced coordinated control method for the reheat steam temperature system of a coal-fired unit according to claim 1 is characterized in that: The calculated flue gas damper opening instruction satisfies the following relationship: CMD fg-min <CMD fg <CMD fg-max Where: CMD fg It is the flue gas damper opening instruction; CMD fg-min The lower limit of the flue gas damper opening instruction; CMD fg-max It is the upper limit of the flue gas damper opening instruction.

Images