JPH0561644B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to a power generation plant control system,
In particular, the present invention relates to a power generation plant predictive optimal control system suitable for keeping the system frequency stable even when the system load demand fluctuates significantly and suppressing fluctuations in the control amount of the power generation plant. The power generation plant control system consists of a central power dispatch center (hereinafter referred to as "intermediate dispatch") that issues load commands to the power generation plant, and a plant control system that controls the power generation plant in accordance with the load commands. Intermediate distribution predicts the load demand of the power system, determines the current load distribution for each power generation plant in the system based on the prediction result, and outputs this current load distribution as a load command to each power generation plant. . Each plant control system controls each power generation plant so that the generator output follows the load command from the intermediate supply. By the way, in a coal-fired power plant, there is a large delay at the operating end (for example, a coal mill), and the heat exchangers are distributed in series, resulting in a large flow delay of the working fluid, so the load command from the intermediate supply may fluctuate significantly. ,
Control variables such as steam pressure and temperature fluctuated greatly, making it difficult to significantly and rapidly change the load command from the intermediate supply. The purpose of the present invention is to keep the grid frequency stable even when the load demand of the power system fluctuates greatly, and as a result, the load command from the intermediate supply fluctuates significantly and rapidly, and to perform good load following control of the power generation plant. The aim is to provide a power plant predictive and optimal control system. The present invention aims to maintain a stable system frequency even when the load demand of the power system fluctuates greatly, and as a result, the load command from the intermediate supply fluctuates significantly and rapidly, and to perform good load following control of the power generation plant. ,
The load demand of the power system is predicted at the intermediate distribution, the current and future load distribution of each power generation plant in the system is calculated based on the prediction result, and this current and future load distribution is used as a load command for each power generation plant. Each plant control system predicts the future movement of the power generation plant in accordance with this load command and controls the power generation plant. An embodiment of the present invention is shown in FIG. Examples are:
It consists of a central power dispatch center 500, a model built-in predictive control system 100, a master controller 200 and a subloop controller 300 of a thermal power plant control system, details of which are shown in FIG. Master controller 2 of thermal power plant control system
00 and subloop controller 300 are
Current load command L _C (=ELD
+AFC) to control the thermal power plant 400 whose details are shown in FIG. The intermediate power supply 500 predicts the load demand of the power system, and based on this prediction result, determines the load distribution of each power generation plant in the system at the current point in time and in the near future (for example, up to 30 minutes in the future),
This current and near future load distribution is output to each power plant as the current load command L _C (=ELD+AFC) and the near future load command L _CP (=ELDP). The predictive control system 100 with a built-in model controls a system that combines a master controller 200, a subloop controller 300, and a thermal power plant 400 of a thermal power plant control system, and controls the system based on the near future load command L _CP from the intermediate supply 500. The model of the controlled object is used to predict the movement of the controlled object in the near future, and the controlled object is optimally controlled based on the prediction result. next,
This will be explained in detail. As shown in FIG. 2, the master controller 200 of the thermal power plant control system determines the water supply flow rate demand based on the load demand L _D obtained by processing the current load command L _C (=ELD+AFC) from the intermediate supply through rate-of-change control. F _FWD , fuel flow rate demand F _FD , air flow rate demand F _AD , spray flow rate demand F _SPD and recirculation gas flow rate demand F _GRD are determined in advance (feed forward control), and main steam pressure P _MS , main steam The demand for each of the above manipulated variables is corrected by feed-back control of the temperature T _MS , gas O ₂ and reheated steam temperature T _RS , and the corrected demand F′ _FWD ,
Create F′ _FD , F′ _AD , F′ _SPD and F′ _GRD . In addition,
Turbine steam flow demand F _MSD is generator output
Determined by MW feed back control. Also, the master controller 200
Rate of change of L _D L _DR and furnace water wall outlet steam temperature T _WW
Find the rate of change T _WWR . As mentioned earlier, the Nakagyo 500 predicts the load demand of the power system, and based on the results of this prediction, determines the load distribution for each power generation plant in the system at the current point in time and in the near future (for example, up to 30 minutes in the future). , and outputs the current and near future load distributions to each power generation plant as the current load command L _C and the near future load command L _CP . Various conventionally known techniques can be used to predict the load demand of the grid, and one example will be described below. In this prediction method, as shown in equation (1), factors that have a large impact on the load are picked up and predicted in the form of a weighted sum. L^(t+τ)=w ₀ +w ₁ x ₁ (t)+w ₂ x ₂ (t)+... +W _a x _a (t)...(1) Where, t: Current time L^(t+τ): Predicted time Predicted value of system load at t + τ x _J (t): Factor at current time t that correlates with system load L (t + τ) at predicted time t + _τ W _J : Weighting coefficient Data and estimated values of temperature, illuminance, humidity, and wind speed at the time of prediction are used, and the weighting of each factor is determined from the actual results during a finite past observation period. The prediction accuracy is around several percent compared to other prediction methods. Next, the ELD prediction calculation based on the system load demand prediction result obtained by equation (1) can basically be formulated as follows. (a) P _Si (τ) is the output share of the i-th thermal power plant, P _Hj (τ) is the output share of the j-th hydroelectric power plant, P _Ak (τ) is the output share of the k-th nuclear power plant, L( τ) is the load demand of the grid, R(τ)
Assuming that is the total amount of loss in the grid, the condition for supply and demand balance can be expressed by the following equation. P _S1 (τ)+…+P _So (τ)+P _H1 (τ)+…+ P _Hn (τ)+P _AI (τ)+…+P _AI (τ)=L(τ) +R(τ) (2) ( b) The upper and lower limits of the output of the i-th thermal power plant are
P ^Max _Si , P ^Min _Si , upper and lower limits of the j-th output of hydroelectric power plant P ^Max _HJ , P ^Min _HJ , nuclear power plant k
Assuming that the upper and lower limits of the th output are P ^Max _Ak and P ^Min _Ak , the constraints on each output are as shown in the following equations. P ^Min _Si P _Si (τ)P ^Max _Si (i=1,2,…,n) P ^Min _HJ P _HJ (τ)P ^Max _HJ (j=1,2,…,m) P ^Min _Ak P _Ak ( τ) P ^Max _Ak (k=1, 2,...,l)
}
(3) (c) Let the water usage rate of the jth hydropower plant be
If Q _J (τ), then P _HJ (τ) is expressed as equation (4).
It becomes a function of Q _J (τ). P _Hj (τ) K(Q _J (τ)) (j=1, 2,..., m) (4) (d) The total water consumption Q _JT in the j-th consideration period T of the hydroelectric power plant is Q _J ( τ) from t ₀ (consideration start time) to t ₀ +T. Operationally, Q _JT is assumed to be given. ∫ ^t ₀ ^+T _t0 Q _J (τ)dτ=Q _JT (j=1,2,…,m)(5
) Note that the ELD prediction calculation is performed over a certain time width (t ₀ ~ _{t 0} +
The consideration start time t ₀ for calculation over T) is
It may be selected after the current time t. (e) The system loss is calculated as {P _S2 }, {P _Hj }, as shown in equation (6).
It is a function of {P _Ak }. R(τ)=R(P _S1 ,…, P _So , P _H1 ,…, P _Hn ,
P _A1 , …, A _AI (6) (f) Using equations (3) to (6) as constraints, the objective function is t ₀ to _{t 0}
+T is the total thermal power and nuclear fuel cost F. The fuel cost per unit time when the output of the i-th thermal power plant is P _S1 is f _Si P _Si ), and the fuel cost per unit time when the output of the k-th nuclear power plant is P _Ak is f _Ak (P _Ak ), the objective function is (7)
It can be expressed as the formula. F=∫ ^t ₀ ^+T _t0 {f _S1 (P _S1 )+...+f _So (P _So ) +f _A1 (P _A1 )+...+f _AI (P _AI )}dτ (7) ELD prediction calculation uses this F like minimizing
P _Si (τ), P _Hj (τ), and P _Ak (τ) are calculated over t ₀ to t ₀ +T (i=1,...n; j=1,
..., m; k=1, ..., l). Static characteristic values are given to the upper and lower limits of the output of the power plant in (b). Further, the fuel cost f _s (P _s ) per unit time of the thermal power plant in (f) varies depending on the model, but qualitatively it is roughly as shown in Figure 4.
As can be seen from the figure, the efficiency changes depending on the output, A
The output at point B is more efficient than when the output is small as at point B, and the fuel cost per unit kW is lower (unit kW when the slope of the straight line is output A)
(This will be the corresponding fuel cost.) However, as the output increases further to C, the efficiency deteriorates. In addition, the fuel cost per unit hour of a nuclear power plant
f _A (P _A ) is also related to plant efficiency, similar to the fuel cost characteristics of a thermal power plant, and is given as a function of output. The medium salary of 500 is P _Si (τ), P _Hj (τ) calculated above.
P _Ak (τ) (t ₀ τt ₀ +T) (i=1,...,n;j
=1,...,m; k=1,...,l) to the current load distribution P _Si (t), P _Hj (t), P _Ak (t) (i=1,...,n;
j=1,...,m; k=1,...,l) at the current time
Output to each power plant as an ELD signal. In addition, the near future load distribution P _Si (τ), P _Hj (τ), P _Ak
(τ) (tτt+T ₀ ) (i=1,...,n; j=
1,...,m; k=1,...,l) in the near future
It is output to each power plant as an ELD signal (=ELD(τ)(tτt+T ₀ )). In addition, T ₀ (T)
can be any length, but it is determined depending on the response characteristics of the power plant. For example, in the case of a thermal power plant, approximately 30 minutes to 1 hour is appropriate. In addition, the intermediate feeder 500 calculates the AFC signal of each power generation plant using the following equation based on the deviation E _f between the current system frequency and the reference frequency _r , and outputs this result to each power generation plant. AFC _i (t)=K _i・( _r −) (8) Here, AFC _i (t): Current power generation plant i
AFC signal K _i : AFC share ratio of power generation plant i The predictive control system 100 with a built-in model is a master controller 20 of a thermal power plant control system.
0, the system that combines the subloop controller 300 and the thermal power plant 400 is controlled, the load demand change rate L _DR from the master controller 200 is used as a disturbance, the main steam temperature deviation T _MSE , the furnace water wall outlet steam Temperature change rate
Using T _WWR and reheat steam temperature deviation T _RSE as control variables, we successively identify a model for the above-mentioned controlled object, and using this model and the near future ELD signal (=ELDP) from intermediate feeder 500, the above-mentioned controlled object The model predicts the movement of the controlled object in the near future, and the quadratic evaluation function is used to optimally control the controlled object. The operation of the model built-in predictive control system is shown in a flow diagram as shown in FIG. This will be explained in detail below. It is assumed that the above controlled object is expressed by an autoregressive moving average (ARMA) model as follows. x(k)=A(1)x(k-1)+A(2)x(k-2)+...
+
A(M)x(k-m) +B(1)u(k-1)+B(2)u(k-2)+...+B
(m)u(k-m) +C(1)v(k-1)+C(2)v(k-2)+...+C
(m)v(k-m) +w(k) (9) Here,

[Table] v (k-l) = [L _DR (k-l)]
(l=1,2,...,M) = [v ₁ (k-l)] T _MSE (k-l): (k-l) Main steam temperature deviation at the sampling time T _WWR (k-l): ( k-l) Rate of change in steam temperature at the furnace water wall outlet at the time of sampling T _RSE (k-l): (k-l) Reheat steam temperature deviation at the time of sampling ΔF _FD (k-l): (k-l) Sampling Fuel flow demand modification signal at time ΔF _SPD (k-l): Spray flow rate demand modification signal at (k-l) sampling time ΔF _GRD (k-l): Recirculation gas flow demand at sampling time Modified signal L _DR (k-l): (k-l) Load demand change at sampling time A(l) = a ₁₁ (l)a ₁₂ (l)a ₁₃ (l) a ₂₁ (l)a ₂₂ (l )a ₂₃ (l) a ₃₁ (l)a ₃₂ (l)a ₃₃ (l): Coefficient B(l)=b ₁₁ (l)b ₁₂ (l)b ₁₃ (l) b ₂₁ (l)b ₂₂ (l)b ₂₃ (l) b ₃₁ (l)b ₃₂ (l)b ₃₃ (l): Coefficient C(l)=c ₁₁ (l) c ₂₁ (l) c ₃₁ (l): Coefficient w(k )=w ₁ (k) w ₂ (k) w ₃ (k): Noise at k sampling time Equation (9) is the observation equation for the coefficient A ₁ (l) (l = 1, 2,..., M). It can be transformed as follows. x(k)=C ₁ (k)A ₁ (k)+w(k) (10) Here, C ₁ (k)=X ^T (k) 0 00 X ^T (k) 00 0 X ^T (k) X(k)=X ₁ (k-1)) X ₁ (k-2) 〓 X1(k-M) X ₁ (k-l)=x ₁ (k-l) x ₂ ((k-l x ₃ (k-l) u ₁ (k-l) u ₂ (k-l) u ₃ (k-l) ₁ (k-l) A ₁ (k)=A ₁₁ (k) A ₁₂ (k) A ₁₃ (k)

[Table] The transition equation for the coefficient A ₁ (k) shall be as follows. A ₁ (k+1)=A ₂ (k) (11) When a Kalman filter is constructed for equations (10) and (11), it is as follows. A^ ₁ (k)=A^ ₁ (k-1) + P(k)C ^T ₁ (k)W ^-1 {x(k)-C ₁ (k)A^ ₁ (k-1)} P( k)=(P ^-1 (k-1)+C ^T ₁ (k)W ^-1 C ₁ (k)) ^-1 (12) Here, A^ ₁ (k): estimated value of A ₁ (k) P(k): Covariance matrix of estimation error of A ₁ (k) W: Covariance matrix of w(k) Substituting A^ ₁ (k) obtained from equation (12) into equation (9), we get the following It becomes like this. x(k)=A^(1)x(k-1)+A^(2)x(k-2)+...
+
A^(M) x(k-M)+B^(1)u(k-1)+B^(2)u(k
-2) +...+B^(M)u(k-M)+C^(1)v(k-1)+
C^(2) v(k-2)+…+C^(M)v(k-M)+w(
)
k) Here, A^(l): Estimated value of A(l) B^(l): Estimated value of B(l) C^(l): Estimated value of C(l) In order to convert into a transition expression, a variable Z ₁ (k) shown in the following equation is introduced. Z _i (k)= _M 〓 ^l=i+1 {A^(l)x(k+i-l)+B^(l) u(k+i-l)+C^(l)v(k+i-l)} (i =1,2,…,M-1) (14) Z ₀ (k)=x(k)= _M 〓 ^l=1 {A^(l)x(k-l)+B^(l) _u u( k−l)+C^(l) v(k−l)}+w(k) When formula (14) is written down, it becomes as follows. Z ₀ (k)=A^(1)Z ₀ (k-1)+B^(1)u(k-1) +C^(1)v(k-1)+Z ₁ (k-1) +w(k ) Z ₁ (k)=A^(2)Z ₀ (k-1)+B^(2)u(k-1)+C^(
2)
v(k-1)+Z ₂ (k-1) (15) Z _M-2 (k)=A^(M-1) Z ₀ (k-1)+B^(M-1) u(k-1 )+C^(M-1)v(k-1) +Z _M-1 (k-1) Z _M-1 (k)=A^(M)Z ₀ (k-1)+B^(M) u( k-1)+C^(M)v(k-1) Equation (15) can be expressed as a state transition expression as follows. Z(k)=Φ・Z(k-1)+P・u(k-1)+H・
v(k-1)+V(k) (16) x(k)=[I0...0]Z(k) (17) Here, Z ^T (k)=[Z ^T ₀ (k)Z ^T ₀ ... Z ^T _M-1 (k)] V ^T (k)=[w ^T (k)0…0] Φ=A^(1) A^(2) 〓 A^(M-1) A^(M) I 0 〓 0 00 I 〓 0 0… … … …0 0 〓 0 0 P ^T = [B ^T (1)B ^T (2)…B ^T (M-1)B ^T (M)] H ^T = [ C ^T (1) C ^T (2)...C ^T (M-1) C ^T (M)] I: Unit matrix The following quadratic form evaluation function is used as the evaluation function τ. J=E∫ _M 〓 ^l=1 {Z ^T (k+i)QZ(k+i) + ^T _u (k+i-1)Ru(k+i-1)}] (18) Here, E: Symbol Q representing expected value: (M・3)×(M・3) order positive semidefinite matrix (weight) R: 3×3 order positive semidefinite matrix (weight) Apply dynamic × programming (DP) to equations (16) and (18) Then, the optimal manipulated variable u ⁰ (k) that minimizes equation (18) can be found using the following recurrence equation.

[Table] From equation (19), u ⁰ (k) becomes as follows. u ⁰ (k)=-B(N) {P ^T・S(N−1)・Φ・Z(k) +P ^T・S(N−1)・H・v(k) +1/2P ^T・θ (N-1)} (20) In other words, equation (20) is the master controller 20
0 to main steam temperature deviation T _MSE , furnace water wall outlet steam temperature change rate T _WWR , reheat steam temperature deviation T _RSE and load demand change rate L _DR , fuel flow rate demand correction signal ΔF _FD , spray flow rate demand correction This is a formula for calculating the optimal values of the signal ΔF _SPD and the recirculation gas flow rate demand correction signal ΔF _GRD . The future value of the load demand change rate _LDR used in equation (19) is the change rate of the ELD signal (=ELDP(τ)(tτt+ _T0 )) in the near future from the intermediate salary shown below.
Use ELDPR. v(k+i) = [L _DR (k+i)] = [ELDPR(k+i)] (i = 1, 2,..., N-1) (21) The subloop controller 300 of the thermal power plant control system is expressed by the equation (22). Modified demand for each manipulated variable shown in F″ _MSD , F″ _FWD , F″ _FD , F″ _SPD and F″ _GR
D
Based on the turbine steam flow rate F _MS , feed water flow rate
F _FW , fuel flow rate F _F , air flow rate F _A , spray flow rate F _SP
and control the recirculation gas flow rate F _GR . F″ _MSD =F _MSD F′ _FWD =F′ _FWD F″ _FD =F′ _FD +ΔF _FD F″〓 _D =F′ _AD +f(ΔF _FD F″ _SPD =F′ _SPD +ΔF _SPD F″ _GRD =F′ _GRD +ΔF _GRD (22) Here, f (ΔF _FD ): Air flow rate demand correction signal corresponding to the fuel flow rate demand correction signal As can be seen from the above explanation, according to an embodiment of the present invention, the power system predicts the load demand of The plant control system predicts the future movement of the power plant based on current and future load commands and controls the power plant, so the load demand of the grid fluctuates significantly, and as a result, the load command from the intermediate supply increases significantly. It is also possible to keep the system frequency stable even when it fluctuates rapidly, and to perform good load following control of the power generation plant.In an embodiment of the invention, the master controller, subloop controller, and thermal power generation The control target is a system that combines plants, and the load demand change rate L _DR
Disturbance, main steam temperature deviation _MSE , furnace water wall outlet steam temperature change rate T _WWR and reheat steam temperature deviation T _RSE are controlled variables, fuel flow rate demand correction signal ΔF _FD , spray flow rate demand correction signal ΔF _SPD and recirculating gas The flow rate demand correction signal ΔF _GRD was used as the manipulated variable, but
Generator output deviation MW _E , main steam pressure deviation P _MSE and gas o ₂ deviation O _2E are added as control variables, and turbine steam flow rate demand correction signal ΔF _MSD , feed water flow rate demand correction signal ΔF _FWD and air flow rate demand are added as control variables. A correction signal ΔF _AD may be added. In the embodiment of the invention, the characteristics of the controlled object are sequentially identified, but it is also possible to identify them once during a test run and then use the identification results during the test run. In the embodiment of the invention, the turbine steam flow rate F _MS and the feed water flow rate are determined based on the modified demands F″ _MSD , F″ _FWD , F″ _FD , F″ _SPD , and F″ _GRD for each manipulated variable shown in equation (22). F _FW , fuel flow rate F _F , air flow rate F _A , spray flow rate F _SP and recirculation gas flow rate F _GR are controlled by subloop controller 300;
The correction demand for each manipulated variable may be determined using equations (23) and (24). F″ _MSD =F _MSD F″ _FWD =F′ _FWD F″ _FD =F′ _FD +ΔF′ _FD F″ _AD =F′ _AD +f(ΔF′ _FD ) F″ _SPD =F′ _SPD +ΔF′ _SPD F″ _GRD = F′ _GRD + ΔF′ _GRD (23) ΔF′ _FD = ΔF _FD + PRS _FD ΔF′ _SPD = ΔF _SPD + PRS _SPD ΔF′ _GRD = ΔF _GRD + PRS _GRD (24) Here, PRS _FD , PRS _SPD , PRS _GRD : Pseudo-random Signal In this case, u(k-l)(l=
1, 2,...,M) are as follows. u(k-l)=ΔF' _FD (k-l) ΔF' _SPD (k-1) ΔF' _GRD (k-1) (24) By doing this, the model coefficients A(l), B
The estimation accuracy of (l) and C(l) can be improved. According to the present invention, the load demand of the power system is predicted in the intermediate supply, the current and future load distribution of each power generation plant in the system is calculated based on the prediction result, and the current and future load distribution is calculated. is output as a load command to each power generation plant, and each plant control system predicts the future movement of the power plant based on the current and future load commands and controls the power plant, so the load demand of the grid fluctuates significantly. As a result, even if the load command from the intermediate supply fluctuates significantly and rapidly, the system frequency can be kept stable and the power generation plant can be properly controlled to follow the load.

[Brief explanation of the drawing]

FIG. 1 and FIG. 2 are diagrams showing one embodiment of the present invention, and FIG. 3 is a diagram showing an example of a thermal power plant.
FIG. 4 is a diagram showing fuel cost characteristics of a thermal power plant, and FIG. 5 is a diagram showing a flow diagram of a model-embedded predictive control system. 201...adder, 202...change rate limiter,
203... Correction circuit, 204... Correction circuit, 20
5... Correction circuit, 206... Correction circuit, 207...
...Correction circuit, 208... Rate of change calculator, 209...
...Subtractor, 210...Main steam pressure controller, 211
...Subtractor, 212 ...Main steam temperature controller, 21
3...Adder, 214...Subtractor, 215...Gas _O2 controller, 216...Main steam temperature controller, 21
7...Adder, 218...Subtractor, 219...Reheat steam temperature controller, 220...Adder, 221...
... Rate of change calculator, 222... Generator output controller,
301... Turbine controller, 302... Water supply flow rate controller, 303... Fuel flow rate controller, 304...
Air flow rate controller, 305...Spray flow rate controller,
306...Recirculation gas flow rate controller, 401...Forced draft fan, 402...Air preheater, 403...1
Air fan, 404... Coal bunker, 405...
...Coal feeder drive motor, 406...Coal feeder, 407
...Coal mill, 408...Secondary superheater, 409...
...Secondary reheater, 410...Furnace water cooling wall, 411...
...Primary superheater, 412...Gas circulation fan, 41
3... Spray control valve, 414... Induced draft fan, 4
15...Water pump, 416...Main steam control valve,
417... High pressure turbine, 418... Medium/low pressure turbine, 419... Generator, 420... Condenser,
421...spray.

Claims (1)

[Claims] 1. A power plant predictive optimal control system that outputs corresponding manipulated variables for at least a plurality of hydroelectric power plants and a plurality of thermal power plants, the system comprising: A power load demand forecasting means that predicts the load demand of the power system based on future environmental conditions and past actual values; and a power load demand forecasting means that predicts the load demand of the power system based on the future environmental conditions and past actual values; a total load calculation means for calculating the total load; calculates the load distribution for each power plant at present and in the future from the fuel cost per unit time for each thermal power plant so that the total fuel cost per unit time for multiple thermal power plants is minimized, A load command output means that outputs each of the load distributions for each plant as a load command; A model that is determined according to the environment of the power plant corresponding to each power plant; and a corresponding power plant at present and in the future. and a manipulated variable calculation means for predicting the future movement of the power plant based on the load command related to the power plant, determining the manipulated variable of the power plant, and outputting the manipulated variable to the power plant. A power generation plant predictive optimal control system characterized by: