JPS6333164B2 - - Google Patents

Description

[Detailed Description of the Invention] The present invention relates to a method for controlling a thermal power plant,
In particular, the present invention relates to a control method suitable for always maintaining plant efficiency at the highest point even when ambient conditions such as seawater temperature and atmospheric temperature fluctuate during normal load operation. Although the issue of highly efficient operational control of thermal power plants has been a hot topic for some time, a control system that aims to improve the efficiency of a plant from an overall perspective has not yet been put into practical use. There are a few reports on optimizing boiler efficiency by manipulating the air-fuel ratio (e.g., F. Moran et al, Development and
application of self-optimizing control to
coal-fired steam-generating plant, Proc.
IEE, Vol. 115, No. 2 (1968-2)). In these conventional control systems, as shown in FIG. 1, a maximum efficiency search is performed in the search means 3 in the control device 2 based on the status feedback 4 from the plant 100 to determine the optimum manipulated variable 5. However, the first problem with the conventional method is that it takes a lot of time to find the maximum efficiency point, that is, the optimum operating amount, because the control is based on status feedback from the plant with a large thermal time constant. The second problem with the conventional method is that it is difficult to actually measure efficiency with high precision, so an efficiency index method is used that uses behavior such as main steam pressure as a substitute for efficiency evaluation. It is a matter of low gender. Furthermore, the third problem with the conventional method is that
Since actual measured values are used, it is susceptible to noise and detection errors. The purpose of the present invention is to eliminate the drawbacks of the conventional methods as described above in a method of controlling a thermal power plant, and to improve the efficiency of the plant from an overall viewpoint. The purpose of the present invention is to provide an efficiency-optimized control method. In order to overcome the problems of the conventional method, the present invention uses a plant model 240 for efficiency calculation built into the control system 200 shown in FIG.
By applying model-based efficiency prediction calculations, we were able to achieve fast and highly accurate efficiency optimization without being constrained by the response speed of the plant. Main control system 20
The basic method of the maximum efficiency search means 230 at 0 is to output a trial operation amount 9 to the plant model 240 before determining the optimum operation amount 11 , and to obtain the corresponding plant efficiency 10 by repeating the procedure. This is a method of searching for the maximum efficiency point. In this case, if the efficiency at the new trial point is higher than the efficiency at the trial point that has already been tried, even if the maximum efficiency point has not been reached, the operation amount corresponding to this new trial point will be used to actually operate the plant. Use as operation command value. FIG. 3 shows the basic configuration of an efficiency optimization control system that is an embodiment of the present invention. However, this diagram only shows functions specific to efficiency optimization, and various minor control systems and plant control systems to which conventional methods can be applied as is are omitted to avoid cluttering the diagram. In FIG. 3, each symbol represents the following. 9a is the trial O ₂ excess rate, 9b is the trial damper opening degree,
9c is the trial condenser vacuum degree, 11a is the optimum O ₂ excess rate, 11b is the optimum damper opening degree, 11c is the optimum condenser vacuum degree, 13a is the economizer (A), and 13b is the economizer.
(B), 14 is a water wall, 15 is a primary superheater, 16 is a secondary superheater, 17 is main steam, 18 is a high pressure turbine, 1
9 is a primary reheater, 20 is a secondary reheater, 21 is reheated steam, 22 is a medium/low pressure turbine, 23 is a generator,
24 is a condenser, 25 is a condensate water, 26 is a low pressure feed water superheater, 27 is a feed water pump, 28 is a high pressure feed water heater,
29 is a circulating water pump, 30 is a parallel damper, 3
1 is a pulverized coal mill, 32 is an air preheater, 33 is a forced draft fan, 34 is a suction draft fan, 47 is a boiler, 5
0 is a turbine, 60 is a water supply system. There are many operating parameters that affect efficiency, but in this example, O ₂ has a relatively large effect.
Three factors were selected: excess ratio m ₁ , parallel damper opening degree m ₂ , and condenser vacuum degree m ₃ , and these were chosen as an example for optimization. The complex method, which is one of the extreme value search methods, is used as the basic algorithm for maximum efficiency search. In the maximum efficiency search means 230 of FIG. 3, two cases where the operating parameters are m ₁ and m ₂ are shown in order to explain the principle of maximum efficiency search using the complex method in an easy-to-understand manner. Now, the trial operation amounts m ¹ ₁ and m ¹ ₂ for trial point 1 are given to the plant model 240 for efficiency calculation, and the efficiency η ¹ as a steady value is determined. Similarly, η ² and η ³ are obtained for trial points 2 and 3. A new trial point 4 is set on the extension of the straight line connecting the point with the lowest efficiency (in this case, 1) and the center of gravity of the remaining points (in this case, on line segments 2 and 3). Then, find the efficiency η ⁴ corresponding to the trial operation amounts m ⁴ _l and m ⁴ ₂ . here,
Similarly, a new trial point 5 is obtained from the newly created triangle 234 excluding trial point 1. At this time, if trial point 5 violates the constraints on the plant state quantity, return to trial point 6 within the domain and create a new triangle 34.
6 is used to determine the trial direction. By repeating this trial method, the maximum efficiency point (7 in this case) can be reached. The manipulated variables corresponding to this point are the optimal values (m ⁷ ₁ , m ⁷ ₂ in this case) and become the actual manipulated outputs for the plant. By continuing such control operations, even if the maximum efficiency point moves due to changes in ambient conditions such as seawater temperature and atmospheric temperature, it is possible to follow this and determine the optimum operation amount. The above has been limited to explaining the principle of efficiency optimization control to which the complex method is applied. Next, an outline of the control algorithm in the embodiment will be explained. FIG. 4 shows the basic processing procedure of efficiency optimization control in the embodiment of the present invention. The control algorithm will be explained step by step below. however,
The symbols are defined as follows. m _i : Manipulated amount i=1 O ₂ excess rate (%) i=2 Parallel damper opening degree (%) i=3 Condenser vacuum degree (mmHg) Here, m ₂ indicates the parallel damper opening degree. However, here it is defined by the following formula. m ₂ = G _S / G _S + G _R ×100 (%) ... (1) Here, G _S is the primary superheater 15,
Represents the gas flow rate in the gas passage where the economizer (A) 13a is installed, and G _R is the primary reheater 19, the economizer
(B) represents the gas flow rate in the gas passage where 13b is arranged. m _{i nax} , m _{i nio} : Manipulated amount upper and lower limits i = 1 O ₂ excess rate upper and lower limits i = 2 Parallel damper opening upper and lower limits i = 3 Condenser vacuum upper and lower limits G _{CL nax} , G _{CL nio} : Upper and lower limits of condenser circulating water flow rate (Kg/sec) D _{T nax} : Upper limit of condenser circulating temperature rise (℃) Y _nax : Upper limit of low pressure turbine exhaust humidity (%) G _{GR nax} , G _{GR nio} : Upper and lower limits of gas recirculation flow rate (Kg/sec) STEP-1 Formation of initial simplex The initial trial points m ¹ _i (i = 1 to 3) shall satisfy all the above constraints, and m ₁ , m ₂ , m ₃ , a polygon (called a simplex) with k angles (k=6 in the case of FIG. 4) is formed, and this is used as an initial simplex. As for this formation method, one point is the initial trial point m ¹ _i , and the remaining (k
-1) points are, for example, uniform random numbers r ^j (j=2 to k)
It is determined by the following formula using . m ^j _i = m _{i nio} + r ^j (m _{i nax} − _{m i nio} ) ...(2) However, 0 < r ^j 1 m ^j _i determined in this way is the constraint condition as the manipulated variable (this is explicitly However, the constraints as process state quantities (referred to as implicit constraints) may not be satisfied. In this case, the trial point is moved in the direction of the center of gravity of the already determined points to the midpoint. In this way all points are suddenly determined. Then, the plant efficiency η ^j (j=1 to 6) corresponding to each point determined through the above trial is calculated as the plant model 240.
Obtained using STEP-2 Calculation of the center of gravity Here, the center of gravity m _Gi of the simplex defined by (k-1) points excluding the point with the lowest efficiency among the points of the simplex is calculated. Now, assuming that the lowest point of efficiency is j=1, m _Gi is expressed by the following formula. Further, the distance Δm _Gi from the lowest efficiency point to the center of gravity is expressed by the following equation. Δm _Gi =m _Gi −m ¹ _i ...(4) STEP-3 Determination of new trial point Set the new trial direction from the lowest efficiency point to the direction of the center of gravity, and move from the center of gravity by α _i times the distance Δm _Gi between both points. Letting the extended point be the new trial point and let it be m ^k+1 _i , it is expressed as m ^k+1 _i = m _Gi + α _i Δm _Gi ……(5). In this case, if an explicit constraint condition is violated, the trial point is set on the constraint condition.
In other words, if the upper limit m _{i nax} is violated, m ^k+1 _i = m _{i nax} ……(6), and if the lower limit m _{i nio} is violated, m ^k+1 _i = m _{i nio} ……( 7). STEP-4 Efficiency calculation Using the plant model 240 for efficiency calculation,
Find the efficiency η ^k+1 corresponding to the new trial point m ^k+1 _i . Here, the simulation range as a plant model covers all systems where energy balance is a problem among systems required for normal load operation. In the turbine system, not only the extraction system but also the seal steam is considered. In addition, in the boiler system, not only heat loss due to exhaust gas but also heat radiation from the boiler wall is considered. Furthermore, the minor control system and plant control system for each aircraft are, of course, simulated. In addition, actual measured values are used for seawater temperature, atmospheric temperature, and wind speed as ambient conditions. STEP-5 Constraint condition monitoring If the process state calculated by the plant model 240 for efficiency calculation violates the implicit constraint condition, all information regarding the trial point m ^k+1 _i is invalidated,
Return to STEP-3 and determine a new trial point. In this case, considering the causal relationship between the operation and the process state, α _i is corrected according to the following equation and the process returns to STEP-3. α ₂ △〓 α ₂ ／2…Condition (G _GR ＞G _{GR nax} )U(G _GR ＜G _GR
nio )……(8) α ₃ △〓 _{α 3} /2… Condition (G _CL > G _{CL nax} ) U (G _CL < G _CL
nio ) U (D _T > D _{T nax} ) U (Y > Y _nax )...(9) Here, α ₁ for the O ₂ excess rate is not corrected.
The reason is that as long as the O ₂ excess rate is operated within the limits, the implicit constraints will not be violated. STEP-6 Convergence judgment Among the efficiencies corresponding to the new trial point and each point constituting the original simplex, calculate the maximum and minimum efficiencies,
Let η _nax and η _nio respectively, and determine whether the maximum efficiency point has been reached according to the following equation. η _nax −η _nio η _nax <ε (10) Here, ε is the criterion for reaching the maximum efficiency point.
If the above formula is satisfied, it can be said that the practical maximum efficiency point has been reached. When the maximum point is reached, the manipulated variable corresponding to η _nax is output, and the process returns to STEP-1 to form an initial simplex. If the maximum point is not reached, the next
Proceed to STEP-7. STEP-7 Determine the direction of efficiency improvement If the new trial point has a higher efficiency than any of the efficiencies of the points that make up the original simplex,
Even if the maximum efficiency point has not been reached, the operation output is
Proceed to the next STEP-8. If there is no operation output,
Proceed directly to the next STEP-8. The reason for providing this function is that if the trial point reliably improves the efficiency compared to the current operating state, then the actual operation output will be performed without waiting for the maximum efficiency point to be reached. STEP-8 Creating a new simplex Among the points that make up the original simplex,
The operating point showing the lowest efficiency is excluded, new trial points are added, a new simplex is formed from the k points, and the process returns to STEP-2. Next, an example of the efficiency optimization control process according to the present invention is shown in FIG. This figure shows the operating parameters.
The trajectory is projected onto three planes in a three-dimensional space defined by the O ₂ excess ratio m ₁ , the parallel damper opening m ₂ , and the condenser vacuum degree m ³ , each moving toward its optimum value. In this figure, 1 to 6 are the vertices of the initial simplex, 7 to 11 are the trial points, the double circle is the maximum efficiency point, and 7
The dotted line connecting points ~11 is the trial operation process, 7~11
The solid line connecting is the optimal operation process. For example, ml
Looking at the relationship between and m ₂ , the optimum point was reached in the 11th trial. However, for m ₃ , although it has come close to the optimal point, it has not reached the optimal point. Although the above has been limited to a general explanation of the control algorithm in the embodiment, the specific method will be additionally explained below. How to determine the initial trial point m ¹ _i and the size of the simplex in the initial simplex formation step 1 is an important issue that affects the convergence of the maximum efficiency search. In order to improve convergence,
We believe that the initial trial point ¹ _i with the highest possible efficiency should be selected, and that the size of the initial simplex should be determined according to this m ¹ _i and the distance to the operational tolerance limit. did. Figure 6 shows the relationship between plant efficiency and O ₂ excess rate m ₁ , and also shows the influence of atmospheric temperature on efficiency characteristics. In other words, the solid line is the atmospheric temperature
The dotted line at 30°C is the efficiency characteristic at 10°C. The upper and lower limits of the permissible range of m ₁ are determined by boiler exhaust gas regulation values, combustion stability, fan capacity, etc., but vary depending on the load level. From this characteristic, the upper limit of the manipulated variable m _{l nax} as a relation to the load L
(L), the lower limit m _{l nio} (L) can be expressed, and the load L
and the manipulated variable m _1M (L, Ta) corresponding to the maximum efficiency point can be expressed as a function of the atmospheric temperature Ta.
Therefore, as shown in FIG. 7, the initial trial point determination unit 96 selects the initial trial point m ¹ _l at the point within the operational tolerance range where maximum efficiency can be expected. Next _, in order to determine the size of the initial simplex according to the distance from m ¹ _l to the operational tolerance limit, m ₁ coordinates m ^j _l (j = ² to k) is determined by the following formula. m ^j _l = m ¹ _l + r ^j (m ¹ _l −m _{l nio} ) K _Ll … (if _l < m ¹ _l ) m ¹ _l + r ^j (m _{l nax} − m ¹ _l ) K _Ll … ( _l m ¹ _l ) ...(11) However, _l = (m _{l nax} + m _{l nio} /2, r ^j is a uniform random number (0 < r ^j 1), and K _Ll is a constant (0 < K _Ll 1). Figure 9 shows the relationship between the plant efficiency and gas recirculation flow rate with respect to the parallel damper opening _m2.The upper limit of _m2 as an allowable range is determined by the gas recirculation fan capacity limit characteristic a and the high-speed gas flow. The lower limit is determined by the undercut prevention limit characteristic b for preventing the overheating of the primary superheater 15 and the economizer (A) 13a due to the
(B) It is determined by the undercut prevention limit characteristic c for 13b, the NOx countermeasure lower limit characteristic d for suppressing the increase in NOx concentration due to the decrease in gas recirculation flow rate, and the reheat steam temperature stabilization limit characteristic e. m _2U2 , m _2U1 , m _2L3 , m _2L2 , m _2L1 are characteristics a, respectively.
These are the limit points determined by b, c, d, and e, and in the case of this figure, the operating range is between m _2L3 and m _2U1 . 8th
The physical basis for the characteristics shown in the figure is that, as is clear from the definition of equation (1), when the value is reduced to _m2 , the amount of gas diverted to the primary reheater side increases, but the reheat steam temperature This is because the control system keeps the reheated steam temperature constant by lowering the gas recirculation flow rate.
The five operating permissible limits shown in Figure 9 change depending on the load level, so the final permissible limit m _{2 nax} and
m _{2 nio} is determined as shown in FIG.
That is, m _{2 nax} is the upper limit m 2U2 obtained by the gas recirculation fan capacity limit calculation means ₁₀₂ and the upper limit m _2U1 obtained by the undercut prevention upper limit calculation means 103.
The lower value among them is determined by the low value selection means 107. Furthermore, m _{2 nio} is the lower limit m _2L3 and NOx
The lower limit m _2L2 obtained by the countermeasure lower limit calculation means 105,
and the lower limit m _2L1 obtained by the reheat steam temperature stabilization lower limit calculation means 106, the largest value is determined by the maximum value selection means 108. Here, m _2U2 ,
m _2L2 , m _2L1 are calculated as functions of load L,
m _2U1 and m _2L3 are calculated as a function of the combustion gas volumetric flow rate V _GAS . Further, V _GAS is calculated as a function of load by the combustion gas volumetric flow rate calculating means 101. As shown in FIG. 9, as m ₂ increases, the gas recirculation flow rate increases, and the power of the gas recirculation fan increases accordingly, so that the plant efficiency loses its droop characteristic. Therefore, assuming that the maximum efficiency point is always at the lower limit of the operational tolerance limit m _{2 nio} , the initial trial point m ¹ ₂ is selected as this point. Also, as in the case of m ₁ , in order to determine the size of the initial simplex according to the distance from m ¹ ₂ to the operational tolerance limit,
m ₂ coordinates of the vertices of the initial simplex except m ¹ ₂ m ^j ₂
(j=2 to k) is determined by the following equation. m ^j ₂ = m ¹ ₂ + r ^j (m ¹ ₂ − m _{2 nio} ) K _L2 … (for ₂ < m ¹ ₂ ) m ¹ ₂ + r ^j (m _{2 nax} − ^{m 1} ₂ ) K _L2 … ( ₂ m ¹ ₂ ) ...(12) However, ₂ = (m _{2 nax} + m _{2 nio} )/2, r ^j is a uniform random number (0 < r ^j 1), k _L2 is a constant (0 < K _L2 1)
It is. FIG. 10 shows the relationship between plant efficiency, circulating water flow rate, and circulating water temperature rise width with respect to condenser vacuum degree _m3 . The physical basis for this characteristic is that increasing m ₃ increases the turbine internal efficiency, but on the other hand, as the circulating water flow rate increases, the pump power increases, and as a result, the plant efficiency is maximized. By having . Furthermore, as the m ₃ decreases, the flow rate of the circulating water decreases, so the temperature rise of the circulating water increases. Here, the allowable operation range is defined by the upper limit characteristic f of the circulating water flow rate determined by the circulating water pump capacity, the lower limit characteristic g of the circulating water flow rate to limit the rate of scale deposition in the cooling pipe, and the lower limit characteristic g of the circulating water flow rate determined by the circulating water pump capacity. This is the upper limit characteristic h of the circulating water temperature rise width for protection. The above limits are determined at the plant planning stage and do not change during operation, but the allowable operating range defined by these limits is greatly influenced by the load level and seawater temperature, as shown in Figure 11. Ru. In the figure, the solid line shows the efficiency when the seawater temperature is 21°C, and the dotted line shows the efficiency when the seawater temperature is 18°C. Therefore, in order to determine the initial trial point m ¹ _3, an initial trial point determination means 10 as shown in FIG.
Use 9. In this means, the upper limit of the manipulated variable is determined as a function of the load L based on the characteristics shown in FIG.
m _{3 nax} (L), the lower limit m _{3 nio} (L), the manipulated variable m _3M (L) corresponding to the maximum efficiency point, and the seawater temperature Tc
A characteristic function corresponding to Tc=T ₁ , T ₂ , ...T _o is prepared with parameters as . Therefore, as shown in FIG. 12, the initial trial point determining means 109:
The initial trial point m ¹ ₃ is selected at the point where maximum efficiency can be expected within the operational tolerance range. Next, in order to determine the size of the initial simplex according to the distance from m ¹ ₃ to the operational tolerance limit, m ₃ coordinates m ³ of the vertices of the initial simplex excluding _m ¹ _l (j = 2 ~ k) is determined by the following formula. m ^j ₃ = m ¹ ₃ + r ^j (m ¹ ₃ − m _{3 nio} ) K _L3 … (for ₃ < m ¹ ₃ ) m ¹ ₃ + r ^j (m _{3 nax} − m ¹ ₃ ) K _L3 … ( ₃ m ¹ ₃ ) ...(13) However, ₃ = (m _{3 nio} + m _{3 nio} )/2, r ^j is a uniform random number (0 < r ^j 1), K _L3 is a constant (0 < K _L3 1)
It is. The various characteristics of the plants shown in FIGS. 6, 8, and 11 can be known using the plant model 240 . In addition, the upper limit of the operation amount determined by each means shown in Fig. 7, Fig. 9, and Fig. 12
m _{i nax} (i=1 to 3) and lower limit m _{i nio} (i=1 to
3) all take into consideration implicit constraints, and have different meanings from the explicit constraints used in equation (2). Therefore, each vertex coordinate m ^j _i (i=1 to 3, j
=2~k) violates the explicit constraint, then
Using the same concept as equations (6) and (7), each vertex coordinate is set under explicit constraints. In addition, the value of the constant K _Li (i = 1 to 3) in equations (11), (12), and (13) can be determined by analyzing the control characteristics through simulation.
What is necessary is to determine an appropriate value that suits the plant to be controlled. In the above description of the embodiments of the present invention, handling of transient states of the plant due to load fluctuations and switching of the number of auxiliary machines in operation has not been mentioned, but this will be explained below. Because plants have a large heat capacity, it is difficult to know their true efficiency during transients. Therefore, in this embodiment, as shown in FIG. 13, the maximum efficiency search means 230 stops operating during the load change and immediately after the load change is completed. In order to detect the presence or absence of load fluctuation, load L(t _{0 )} , L(t ₀ −
Δt), ...L(t ₀ −nΔt), the maximum value is L _nax and the minimum value is L _nio , and if L _nax −L _nio >ε _L ...(14), it is considered that there is a load fluctuation. . That is, until time _t2 in FIG. 13, the load is regarded as steady according to equation (14), and after detecting load fluctuation at _t2 , the maximum efficiency search means 230 is stopped. Even after the load change is completed, the maximum efficiency search means 230 remains inactive until time _t4 when the plant is in a thermally transient state. After t ₄ , it is considered to be in a steady state and operates again. In this case, the maximum efficiency searching means 230 starts its operation from the initial simplex formation step 1 shown in FIG. Here, as the value of nΔt, a necessary and sufficient value is used in consideration of the transient characteristics of the plant to be controlled. When there is a large load change that involves a change in the number of auxiliary machines in operation, such as a pulverized coal mill or water pump, the maximum efficiency search means 230 is activated or deactivated in consideration of the thermal transient state caused by the change in the number of auxiliary machines. let
In other words, as shown in Fig. 14, when the example shown in Fig. 13 involves switching the number of auxiliary machines in operation, the heat equilibration time t _S required during the number switching sequence and as a result of the number switching has elapsed. After that, the maximum efficiency search means 230 is operated. If the number of auxiliary machines in operation is switched even in a steady load state, the operation will be stopped only during the switching sequence and during t _S. FIG. 15 shows the processing procedure of the embodiment for the above purpose. This process operates with a period Δt,
The sequence operation determination means 111 determines whether the sequence for switching the number of auxiliary machines in operation is in operation, and if it is in operation, the timer reset means 11
2, the timer is reset to measure the time t _A after the completion of the switching sequence, and the maximum efficiency search means 2
30 is stopped by the stop command means 113. On the other hand, when the switching sequence is completed, the timer 114 measures the subsequent elapsed time _tA .
Next, the load fluctuation state determination means 115 determines whether or not the load is fluctuating according to equation (14). If the load is changing, the operation of the maximum efficiency searching means 230 is stopped, and if the load changing is completed, the thermal equilibrium state determining means 116 determines the thermal equilibrium state by comparing the magnitudes of _tA and _tS . If the state is in thermal equilibrium, the operation command means 117 causes the maximum efficiency search means 230 to operate. In this embodiment, when determining the coordinates of each vertex of the initial simplex, the initial trial point m ¹ _l regarding the O ₂ excess rate is determined by considering both the load L and the atmospheric temperature Ta. However, since the influence of atmospheric temperature on plant characteristics is relatively small, it is not necessarily necessary to take atmospheric temperature into consideration, and even if the method determines the initial trial point m ¹ _l by simply considering only the load, the control characteristics will be significantly deteriorated. There's no need to worry about it. In addition, as shown in Fig. 4, when the manipulated variable converges to the maximum efficiency point, the initial simplex is again formed by the method shown in Figs. 6 to 12. As long as the operation of the maximum efficiency search means 230 is not interrupted due to a change in the number of machines in operation, the control characteristics will not be impaired even if the initial simplex is formed near the convergence point. In addition, in this example, the direction of the new trial point is
Although the direction of the center of gravity expressed by equation (3) is determined from the vertex of the simplex where the efficiency is lowest, it is not necessary to follow this direction, and stable efficiency optimization can also be achieved by the following method. It is possible. The first method is to create a new trial on a straight line that passes through the centroids of the p points with the lowest efficiency and the centroids of the remaining (k-p) points among the k vertices forming the simplex. This is a method of determining points. In addition, in the second method, k vertices are replaced by A, which consists of p vertices with lower efficiency.
A method of determining L new trial points in a straight line direction passing through the center of gravity excluding the l (l = 1 - p)th point from the lowest efficiency point of both groups. It is. In addition, in the third method, when calculating the center of gravity m _Gi of the simplex, instead of using equation (3) to directly use the efficiency η ^j _i of each vertex as a weighting coefficient, the difference η ^j _i − This method is calculated using the following equation, where is the weighting coefficient. However, it is desirable to modify it according to the load level. In the first and second methods described above, stable convergence can be expected without being affected by the singularity of the lowest efficiency point. In addition, compared to equation (3), the third method based on equation (15) more accurately determines the direction of maximum slope of the efficiency characteristic with respect to the manipulated variable and determines a new trial point, resulting in better convergence. can be expected. In addition, in this embodiment, if a new trial point has a higher efficiency than any of the efficiencies of the points that make up the original simplex, the operation output can be performed even if the maximum efficiency point has not been reached. Avoid sudden changes in volume. Furthermore, in order to increase this effect, it is also effective to select n vertices of the simplex from those with the highest efficiency, and to use their center of gravity as the actual manipulated variable. In this case, it is effective to select the value of n within the range of 2nk/2. In addition, in this example, if the process state at the new trial point violates the implicit constraint, only the manipulated variables directly related to the violated constraint are calculated according to equations (8) and (9). It's being pushed back. but,
Since the causal relationship between the process state to which constraints are imposed and the amount of manipulation does not necessarily correspond one-to-one,
Assuming that equations (8) and (9) are used as is, α _i is corrected according to α _i △〓α _i (1−β) ……(16) for manipulated variables that are not directly related. desirable. However, in the above equation, β may be approximately 0.2. Another effective method is to estimate the operation amount limit based on the degree of violation of the implicit constraint condition and to retreat to the limit value. This method is effective for optimizing the manipulated variable whose efficiency characteristic has a monotonous relationship with the manipulated variable. Now, if the state quantity at the lowest efficiency point in the simplex is x ₁ , the state quantity at the new trial point is x ₂ , and the constraint condition is x _L , then in order to correct α and retreat the manipulated variable to the limit value, , for example, α' can be determined by linear interpolation using the following equation: x ₂ −x _l /1+α=x _L −x _l /1+α′ (17). In reality, the correction coefficient ξ(0<ξ<
Using 1), a new trial point is determined using the following formula. α′={(1+α)x _L −x _l /x ₂ −x _l −1}ξ……(18) In addition, in this example, the maximum efficiency search is performed according to the load fluctuation state and the switching of the number of auxiliary machines in operation. Means 23
0 is activated or paused, but in this case,
Load fluctuation monitoring section nΔt shown in Figures 13 and 14
does not necessarily have to be constant; in short, it is sufficient that it is the minimum necessary time for the plant to reach a state of thermal equilibrium. Therefore, by sequentially correcting the load level and load fluctuation width, it is possible to increase the chances that the maximum efficiency search means 230 operates.
In a thermal power plant, the thermal time constant is generally smaller during high-load operation than during low-load operation, so the heat equilibrium time becomes shorter during high-load operation, and nΔt can be made smaller. Furthermore, it is desirable to modify the time t _S required for heat equilibrium due to switching the number of operating auxiliary machines according to the load level for the above-mentioned reasons. This _tS is further
It also varies depending on the type of auxiliary equipment whose number is to be changed, so it is desirable to make corrections in consideration of the load level and the type of auxiliary equipment. This makes it possible to increase the operating rate of the efficiency optimization control means as much as possible,
The degree of contribution to highly efficient plant operation can be improved. The first effect of the present invention is that because predictive control is performed using a plant model built into the control system, efficiency optimization can be achieved quickly within 5 minutes without being constrained by the response speed of a plant with a large thermal time constant. This makes it possible to significantly improve the operating rate of the optimization function. The conventional method requires 30 to 60 minutes for optimization, which means that the optimization function cannot be achieved unless the steady load condition continues for any longer. It can be said that it can hardly be put to practical use for operational needs. The second effect of the present invention is that by optimizing the manipulated variables of multiple operating parameters that affect plant efficiency, efficiency can be improved from a comprehensive perspective of the plant, and the conventional method targeted at improving the efficiency of individual equipment. The point is that the efficiency can be significantly improved compared to the conventional method. The third effect of the present invention is that when optimizing efficiency, the plant efficiency directly calculated using the plant model built into the control system is used, so the conventional method of indirectly detecting efficiency based on the efficiency index method is used. The point is that the accuracy of convergence to the optimal value is higher than that of . The fourth effect of the present invention is that since efficiency is optimized using a plant model built into the control system, it is more susceptible to noise and detection errors than conventional methods that optimize efficiency based on actual measured values. figure,
This makes it possible to perform stable and highly accurate optimization. The fifth effect of the present invention is that by preventing sudden changes in the manipulated variable, it is possible to operate the plant with high safety and stability.

[Brief explanation of drawings]

Figure 1 is for explaining the structural differences between the conventional control system and the control system of the present invention shown in Figure 2. FIG. 3 shows the basic configuration of an efficiency-optimizing control system for explaining the structural features of the control system of the present invention in comparison with the control system of FIG. FIG. 4 shows the basic processing of efficiency optimization control. FIG. 5 shows an example of the efficiency optimization control process. FIG. 6 shows the characteristics of plant efficiency versus O ₂ excess rate. FIG. 7 shows a method for determining the initial trial point and operating amount limit for the O ₂ excess rate based on the characteristics shown in FIG. FIG. 8 shows the characteristics of plant efficiency and gas recirculation flow rate with respect to parallel damper opening. FIG. 9 shows a method for determining the initial trial point and operation amount limit for the parallel damper opening based on the characteristics shown in FIG. FIG. 10 shows the characteristics of plant efficiency, circulating water flow rate, and circulating water temperature rise width with respect to the degree of vacuum of the condenser. 1st
Figure 1 shows the characteristics of plant efficiency with respect to condenser vacuum degree using load level and seawater temperature as parameters. FIG. 12 shows a method for determining the initial trial point and operating amount limit for the condenser vacuum degree based on the characteristics shown in FIG. 11. FIG. 13 shows the operating section and the stopping section of the maximum efficiency search means due to load fluctuations. FIG. 14 shows the operation section and the stop section of the maximum efficiency search means due to load fluctuations and switching of the number of operating auxiliary machines. FIG. 15 shows processing means for realizing the purpose shown in FIG. 14. DESCRIPTION OF SYMBOLS 100 ...Plant, 2...Control system, 3...Maximum efficiency search means, 4...Status feedback, 5...
Optimal operation amount, 200 ...Control system, 230 ...Maximum efficiency search means, 240 ...Plant model, 9 ...
Trial operation amount, 10...Plant efficiency, 11 ...Optimum operation amount.

Claims (1)

[Claims] 1. A plant model that calculates plant efficiency as a function of the manipulated variable of a thermal power plant in which at least one of the O ₂ excess rate, parallel damper opening, and condenser vacuum is used as an operational parameter for operational control; comprising a means for modifying the manipulated variable of the parameter according to the plant efficiency obtained through a trial operation using the plant model; and by repeatedly using the plant model and the manipulated variable modifying means, the In a control method that searches for the manipulated variable that maximizes plant efficiency,
During the maximum efficiency search process, if the efficiency at a new trial point is higher than the efficiency at an already tried trial point, the plant is actually operated with the amount of operation corresponding to this new trial point, even if the maximum efficiency point has not been reached. An efficiency optimization control method for a thermal power plant, characterized in that an operation command value is set for a thermal power plant.