WO1996029771A1

WO1996029771A1 - Process for the sequential pilot-control of a process

Info

Publication number: WO1996029771A1
Application number: PCT/DE1996/000545
Authority: WO
Inventors: Jens Albrecht; Harro Kiendl
Original assignee: Vew Energie Ag
Priority date: 1995-03-22
Filing date: 1996-03-22
Publication date: 1996-09-26
Also published as: DE19510343C2; DE19510343A1

Abstract

The invention concerns a process for optimizing the actual charge of a power station by means of a power station control system. According to the invention, a nominal output value for a group of power stations is divided in an optimum manner into nominal values for the individual power stations, taking account of restrictions concerning the individual nominal values, secondary conditions and objectives concerning the power station control system as a whole. This object is achieved by a foresighted process based on a cost-dependent rating function which provides a functional value for all permissible output divisions in dependence on the previous output divisions and a penalizing function if cumulative and reserve requirements are not met.

Description

"Process for sequential feedforward control of a process"

The invention relates to a method for sequential pilot control of a process according to the preamble of claim 1. It relates in particular to a method for optimizing the current use of a power plant by a power plant management system. Power plants are all power generation units that can be influenced in their performance by specifying an active power setpoint.

A process to be controlled P of the aforementioned type consists of n parallel sub-processes P ₁ , P ₂ , ... P _n with input variables u ₁ , u ₂ , ... u _n and output variables y ₁ , y ₂ , ... y _n . For the input variables there are individual restrictions of the form u _i , min ≤ u _i ≤ u _{1, max} and v _{i, min} ≤ v _i = du _i / dt ≤ v _i , _max . This results in the distances Sum u _i and ∑ v _i from the summary limits summ u _{i, min} , ∑ u _{i, max} and ∑ v _i , _min , ∑ v _{i, max} , which are the total reserves for the sums Sum u _i and ∑ v _i . The output quantity y of the process P arises from the individual restrictions in the form y = y ₁ + y ₂ + ... y _n , and the process operation is determined by a cost function f, which is the sum of partial cost functions f _i (u _i ) results, evaluated. In order to achieve a desired output size profile y _A (t), the subprocesses P _{i are} subjected to input size profiles u _i (t. These profiles u _i (t) are obtained from the Sur_τe of the components U _{i, A} (t) and u _{1, B} (t ), which are generated by two separate devices - namely a control A and a controller B. The controller B reacts to a specified controlled variable eg, which arises from deviations of the output variable y from the desired output variable profile Y _A (t) ^and from disturbances. A future setpoint curve e _A (t), which generates the output variable curve Y _A (t), is used for pilot control A.

The goals of power plant management systems that only run thermal power plants can be described as follows: Power plant management systems are installed in the network control systems of energy supply companies to automate the use of the power plants in their own supply area. Power setpoints are generated for all power plants to be operated, which are then transferred to the power plants.

The German laid-open specification 41 35 803 describes a distributed computer system for impressing special operations, in particular start-stop commands, on the power plant processes. The known control device represents a system for the joint operation and joint control of a power plant, which consists of a large number of individual power plants. According to this, the power plant should be able to be operated by simple start, stop and load operation commands, structured by a central monitoring control device. However, the operating conditions to be taken into account here according to the exemplary embodiment (FIG. 4) only concern operating commands to the individual power plants, but not the optimization according to the invention.

The power plant management systems perform other sweeping tasks: - Ensuring a balanced active power balance (power balance) between generated and consumed power in the own supply area, so that no unwanted power flows between the network partners occur;

- Coordination and optimization of the power plant use, so that the power plants are operated economically and gently;

- Establishment of a balanced current account in the medium term in order to keep the mean value of the unwanted power flows between the network partners low.

To do the job, power plant management systems include the following key components: - A module for estimating the current load and predicting the load development; - a module for regulating (regulation module) the active power balance in one's own supply area; - A module for coordinating and optimizing (optimization module) the current power plant use.

For power plant use, the operating mode determines which power plants are controlled by assigning a power setpoint. The optimization module and control module deliver a power setpoint for each of these power plants. The setpoints are transmitted to the power plants either individually or as total setpoints.

The way in which power plant management systems work is subject to the following restrictions and additional conditions: - The power setpoints for all power plants are limited upwards and downwards. The performance limits must not be violated by the target values. There is also a maximum positive and a maximum negative limit for the gradients of the power setpoints (maximum power changes; FIG. 1). These restrictions are usually different for each power plant.

- The optimization module contains a total setpoint as a default. Unless there are any restrictions to the contrary and it makes sense, taking into account the forecast changes in time, this total setpoint should be completely distributed to the power plants by the optimization module, so that the sum of the individual setpoints results in the total setpoint. - The optimization module is also intended to ensure that the control module can restore the balanced active power balance quickly at any time. The manipulated variable reserves should be symmetrically around zero, so that regardless of the sign, a disturbance in the active power balance can always be corrected (Fig. 4). The size of the desired and automatically available reserve (chne Consideration of reserve contracts) is based on the size of the expected disruptions (standard expectations).

German laid-open specification 43 15 317 describes a guide device for the automatic control of power plant blocks in an electrical power supply system, in particular by means of measures to increase performance and frequency support. The control is intended to maintain a predetermined level of reliability of the energy and power supply and to avoid an unnecessarily high reserve power.

In the known according to German Offenlegungsschrift 43 15 317, a guide device is therefore proposed which comprises a first control unit, which is connected via a data transmission element to each power station block to be controlled, and which receives block-specific measured variables and outputs block-specific reference variables. The block-specific measured variables Mn include actual values of the block power and the live steam pressure, as well as valve positions and temperatures measured in or on a turbine housing. In the block-specific measurement variable. Constant block-specific sizes and specifications are taken into account.

Conventional methods for optimizing the use of power plants work on the basis of - generally quadratic - cost functions for each power plant. Depending on the power setpoint of the power plant, these cost functions approximate the fuel costs incurred in the power plant for generating the power. The power limits of the power plants limit the. Search space in which all permissible power setpoints are located; the maximum permissible setpoint gradients of the power plants and the setpoint value changes that are possible based on the current power setpoint, as shown in FIG. 6 for the case of two power plants, restrict the search area S ₁ to the search space S ₂ that can be reached.

In order to ensure manipulated variable reserves for control at any time, fixed control bands are defined for selected power plants (Fig. 2). These rule bands tighten the restrictions for the optimization module by subtracting the associated fixed rule bands from the power limits of the power plants (lowering the upper and raising the lower power limits), so that the search space S _{2 is} reduced (see FIG. 7). In total, a performance band is reserved for the sole use of the control module.

In addition, depending on the process variant, the restrictions for the optimization module are further tightened by limiting the maximum permissible setpoint gradients of the power plants, to which a fixed set of rules is assigned, for optimization to a fixed proportion of the permissible setpoint gradient. This ensures that the control module can use at least the remaining portion of the setpoint gradient without violating the maximum permissible setpoint gradient. All in all, a summary setpoint gradient is reserved for the sole use of the control module. This means that the regulation of the active power can be quickly compensated for by the control module within the defined reserve at any time. The starting point for the optimization is the current power distribution. In the accessible part of the search space, the optimization on the level on which all the points that meet the predetermined total setpoint lie, is the point that leads to the lowest total costs. This point is the starting point for the next optimal division of performance.

Different optimization algorithms are used to solve this optimization task, primarily the step-by-step algorithm (a simplified step-by-step gradient method), or the so-called Euler-Lagrange method (global optimization method); see. see "Elektrizitätswirtschaft", 1970, 327 to 332; "etz-a", 1978, 416 to 421.

According to one finding on which the invention is based, the conventional optimization methods have the disadvantage, due to the structure, that the optimization results are currently not optimal on average over the following reasons:

- The tightening of the restrictions caused by the gradient limitation described above forces the desired gradient reserves to always be reserved for the control module, even if the optimization module actually needed these reserves instead to meet the sum requirement. - The degree of freedom of the optimization are restricted by this tightening of the restrictions even if they are free for the regulation total reserve held by the regulation is currently not required. - The actually available and therefore usable manipulated variable reserve depends heavily on the power distribution by the optimization module and can be larger than required due to the tightening of the restrictions. If the regulation module is limited to the use of the required reserves, such an over-fulfillment of the reserve requirements is of no advantage. - The reserve of power in the power plants is made available by selecting those power plants to which a fixed set of rules is assigned and is not automatically adapted to the current power distribution. In particular, there is no guarantee that the manipulated variable reserve is currently available for optimization purposes. - The provision of manipulated variable reserves generally causes additional costs, so that the cost-optimal allocation of services without cheaper restrictions is cheaper. With conventional methods, it is not possible to dynamically allocate the reserves to the current supply situation. adjust that additional cost by providing before. Reserves are only actually created when needed (if the reserve is actually needed). - The current optimization of the current power plant use for the next time step (momentary optimization) does not necessarily lead to an optimal use of the power plant in the longer term, because the forecast load development has no influence on the momentary optimization. Because of the limited setpoint gradients, this can result in the later required total setpoints not being achieved or in the fact that the later power distributions are not cost-optimal. - Additional information about previously known changes to timetables, target values, actual values, performance limits or operating modes of individual power plants are not used.

These shortcomings cannot be remedied with the conventional optimization methods, because the manipulated variable reserves are assigned to individual power plants after a previous selection and a component for taking predicted future changes into account is missing.

The invention is therefore based on the problem of creating an optimization method for optimally dividing a power setpoint, in particular for a power plant park, to determine setpoints for the individual power plants, taking into account restrictions with respect to the individual setpoints and secondary conditions, as well as target concepts (e.g. with regard to the power plant management as a whole) ; The goal of optimization is to find the power split, with both all constraints and restrictions. are met and the total value of all cost functions becomes minimal and the optimization results are optimal at the moment as well as on average over time.

The solution according to the invention is described for the above type in the characterizing part of claim 1. Improvements and further refinements of the invention are specified in the remaining claims.

In the method for sequential feedforward control of a process mentioned at the outset, two cases are distinguished according to the invention: a) In the event that the next setpoint e ^ for feedforward control A cannot be completely divided into the input variables u _{i, A} within a cycle time without the maximum possible input variables u _{i, A are} selected in total with the current input variables u _{i, B} starting from predefined input variables u ' _{i and} the next input variables u _i violating the individual restrictions. b) For the other case, for which the setpoint e _{A of} the pilot control A can be completely divided into the input variables u _i, A in one or more ways, a distribution is selected for which there is a total reserve for the control of the controller B. is. This division is determined in such a way that within a predetermined value range [u _{B, min} ; u _{B, max]} for the sum of the input variables u _{i, B,} starting from the current sum Σ u _{i, B,} with constant input variables u _{i, A} and in compliance with the individual limitations of the input Great _I the sum of the input variables u _{i, B} with the minimum Rate of change v _{B, min reaches} the value u _{B, min} and with the maximum rate of change v _{B, max} the value u _{B, max} and in addition the cost function f for the input variable distribution u _{i, A is} minimal.

According to the invention, therefore, for the purpose of optimization, a power plant park's power distribution is to be found in which both all constraints and restrictions are met and the total value of all cost functions is minimal, and the optimization results are optimal at the moment and on average over time.

The most important tasks of the guiding device (direction) are - to define an operational plan and a reserve plan for primary control for each power plant block, - to determine dead time and rate of change for a block-specific minute reserve from block-specific measured variables and constant variables.

The method according to the invention can be used to solve the following generalized problem:

Given a process P, which consists of n parallel. Partial processes P _i with input variables u ₁ , u ₂ , ... u _n and output variables y ₁ y ₂ ,. , , y _n exists. For the input variables u _i , individual restrictions of the form u _{i, min} ≤ u _i ≤ u _{i, max}

given. The input variables are composed of two parts u _{i, A} and u _{i, B} , which add up to the input variables u _i . The output variable y of the process P arises from the sum of the output variables y _{i of} all subprocesses P _i ; see. Fig. 15.

A given controller B, which essentially reacts to deviations of the output variable y from a desired output variable profile y _A (t) and to disturbances, generates the input variables u _{i, B.}

A pilot control A is to generate the input quantities u _{i, A} for the subprocesses P _i such that firstly the output quantity of the process follows the output quantity curve y _A (t) and secondly a cost function f which evaluates the operation of the process becomes minimal, and thirdly A reserve for the input quantities u _{i, B} from the control system is kept available so that the control system can intervene quickly in the event of a deviation or malfunction without violating the individual restrictions.

A main application of the optimization method according to the invention is to optimize the current power plant use in a power plant management system. The method according to the invention can be used cyclically, acyclically or cyclically with acyclic processing requirements. The optimal power distribution, which is to be produced according to the cycle or the course of the optimization process, is determined under the restrictions and additional conditions described. The method according to the invention has the special feature, particularly when used for optimizing the use of power plants, that power reserve requirements are optimally met economically and predictable changes in input variables over time can be taken into account in advance when optimizing.

In contrast, in the prior art, e.g. according to the German published specification 43 15 317 discussed above, the primary regulation and reserve keeping take place decentrally in the power plants; there is no central pilot control or optimization of the power distribution to a power plant park. In the known, among other things, the following essential determinations: - two separate devices A and B (pilot control device and controller), - a cost function f to be minimized, which results from partial cost functions,

- Input variables u _i , which are formed from the sum of two parts, - Summary reserves for the sums Σ u _i and Σ v _i .

The invention is explained below with reference to the accompanying drawing in the event of an optimization of the current power plant use of the closer. The drawing shows:

Fig. 1 the Scllwert Limiter. u _{i, min} and u _{i, max} one

Power plant and possible setpoint changes; FIG. 2 shows the setpoint limits of the optimization according to FIG. 1 with reserved control band RB _i ;

Fig. 3 shows the setpoint limits of the optimization

1 with permissible control band ranges;

4 shows the summary control band of a power plant park;

FIG. 5 extended reserve conditions for FIG. 4;

6 shows a search space S ₁ restricted to the accessible search space S ₂ in two power plants;

Fig. 7 search spaces S ₁ , S ₂ with fixed control bands for both power plants;

8 shows a predictive restriction of the search space S ₂ that can be reached to S ₃ ;

9 shows the sum of a current power distribution

P ₀ and predicted total setpoints E ₁ , E ₂ ,

E ₃ ;

10 shows a path P ₀ - P ₁ - P ₂ - P ₃ in the search space S ₁ with optimized power distributions P ₁ , P ₂ and

P ₃ ;

Fig. 11, the web P ₀ - P ₁ - P ₂ - P ₃ in the search space S ₁ at exuberant sum setpoint E _2;

Fig. 12, the web P ₀ - P ₁ - P ₂ - P ₃ in the search space S ₁ at exuberant sum setpoint E _1; Fig. 13, the web P ₀ - P ₁ - P ₂ - P ₃ in the search space S ₁ at omitted total set values E ₁ and E _2;

14 shows a path optimization from A to C ₁ or C ₂

B at t _AB > t _BC and

15 shows a process P with pilot control A and control

B.

1 to 3, the setpoint limits (power setpoint) of an individual power plant are defined as a function of time t. The upper and lower power limits u _{i, max} and u _{i, mix are shown} as well as the maximum positive power change (maximum positive gradient) and the maximum negative power change (maximum negative gradient), each based on a current power setpoint u _i (O).

In an individual power plant, a control band RB _{i is} additionally reserved at the upper and the lower power limit in accordance with FIG. 2, in order to ensure manipulated variable reserves for regulation at all times. These rule bands tighten the restrictions for the optimization by restricting the lower and upper performance limits for the optimization u _{i, A, min} and u _{i, A, max} .

In Fig. 3, the limit of the reserved control band is drawn in dashed lines, because a feature of the invention is that the restricted performance limit for the optimization within a control band, which is generally applicable to the regulated power plants of a power plant park, is permeable for the individual power plant. are. 4 and 5 relate to the representation of a summary rule band, with the total power in the ordinate and the time t in turn in the abscissa. The control bands of the entirety of the power plant fleet are fully available for control, based on a current total control power. Within the control band, as shown in FIG. 5, according to the invention a minimum control power change upwards and a minimum control power change downwards are required. The change in performance should have its minimum size until the upper or lower limit of the control band is reached.

Fig. 6 shows the search area S ₁ in the case of two power plants. Power 1 of one power plant is plotted on the abscissa, power 2 of the other power plant is plotted on the ordinate. The search area S _{1 is shown} , in which there are valid setpoints for both power plants which do not violate the performance limits. Based on the current power distribution P _{0 of} the total power E ₀ between the two power plants, the search space S ₂ that can be reached within S _{1 is} shown, which contains valid setpoints after the optimization for both power plants, which additionally do not violate the maximum power changes. Starting from the current power distribution P ₀ , levels of constant total power E ₀ , E ₁ and E _{2 are} shown schematically.

FIG. 7 shows the search space S1 and the achievable search space S ₂ for fixed control bands RB ₁ , RB ₂ for both power plants, the power of the power plant 1 being shown on the abscissa and the power of the power plant 2 being on the ordinate is. Similarly, the predictive limitation of the search space S ₂ to S ₃ that can be reached in FIG. 8 is for the case schematically shown that the setpoint change of the power plants involved, which is directed against the tendency of the total setpoint, is prohibited depending on the size of the total setpoint change. FIG. 9 shows an example of total setpoints on the ordinate as a function of time (abscissa).

9 is based on the fact that, based on the sum of the current power distribution P _0, total setpoints E ₁ , E ₂ and E3 are predicted, for example for the next few minutes. If the next optimum power distribution is determined from a power distribution to the next forecast value after a time interval in the power representation of FIG. 8, the path P ₀ , P ₁ , P ₂ , P ₃ shown in FIG. 10 results in the search space with optimized ones Power distribution at the levels of constant total power E ₀ to E ₃ .

11 to 13, a finite number of paths can be generated as a combination if the power distributions are not optimized as before for intermediate forecast values. The total setpoint E _{2 is} omitted in FIG. 11, the total setpoint E ₁ in FIG. 12 and the total setpoints E ₁ and E ₂ in FIG. 13. The levels of constant total power are drawn in dashed lines. This is discussed in more detail below.

Substitute power allocations that have not been optimized as before can be selected with the total setpoint so that the path length from the previous to the next is minimal via this substitute power part. In the event that there are different times between the previous, the current and the next power distribution, the length of the tracks is determined by the sum of the products of the individual lengths of the sections with the reciprocal of the associated time differences according to FIG. 14. The drawing shows the path optimization from A to C ₁ or C ₂ via B in the event that t _AB > t _BC .

15 relates to the process P with a pilot control A and a control or a controller B. The process P consists of subprocesses P ₁ , P ₂ , ... P _n . The pilot control A with the setpoint e _A supplies input quantities u _{i, A} as part of the input quantities u _λ for the subprocesses P _i (i = 1,2 ... n). The controller B, to which the setpoint value eg is assigned, supplies input variables u _{i, B} as part of the input variables u _{i mentioned} . The output variables y _{i of} the sub-processes P _i are combined to form the output variables y (t).

The following strategy elements are to be used in the method according to the invention:

A. Starting point for the division of benefits

According to the invention, the starting point for determining the optimal power distribution for the next time step is either the current target value distribution, the current distribution of the existing actual power levels or a combination of jealous divisions.

B. Extended summary reserve conditions. In addition to a summary rule band (power reserve) for balancing the active power balance through the control, another summary reserve condition is used according to the invention:

Within the summary control range, a minimum control speed is to be guaranteed for the control for the time required to reach the control range limits with this minimum control speed, starting from the current position in the control range (required possible control power change within the control time). Before the control band limits are reached, this minimum control speed should not be undercut on average (FIG. 5). The additional summary reserve condition applies equally to the regulation upwards and downwards. The control times are therefore dependent on the current reserve usage by the control module. This strategic element means waiving individual gradient reserves for the regulation by the individual power plants. This leads to a dynamic manipulated variable reserve being available to the control according to the invention.

C. Extended Rulers

Optionally, according to the invention, a maximum permissible power component for the control can be assigned to each power plant instead of a fixed control band. This is not a fixed assignment of rule bands. awarded; for each power plant there is instead a permissible range of rules (Fig. 3. Power plants that are not involved in the control are not assigned a permissible power component for the control. This additional flexibility can be can be used according to the optimization: Depending on the current total setpoint and the associated cost-optimal power distribution to the power plants, the required reserves are available in an optimally distributed manner.

D. Summary reserve estimate

In order to determine which total reserve actually exists for a given power distribution, according to the invention the setpoint changes possible in both directions in the expanded power plant control bands per power plant are first estimated. These are the setpoint changes that are possible within the control times, based on the current power setpoint. Estimation models can also be used to estimate the power plant changes in power expected during these setpoint changes within the standard times.

The summary reserve results from the above from the sum of the possible setpoint changes (manipulated variable reserve) or the expected power changes (power reserve) of all power plants. This makes it possible to estimate how large the average position or performance reserve for a certain power distribution is over time.

E. Extended cost function

Oscillating or strongly fluctuating sol value curves for individual power plants are undesirable because they cause high material stress. In order to take this into account, the cost functions according to the invention are terms for punishing target values rungs and (especially discontinuous) gradient changes expanded. In addition to the individual cost functions of the power plants, a cost function for the unwanted exchange of services between the network partners, triggered by a missing total output in their own supply area, is supplemented.

F. Punitive function for failure to comply with the

Total and reserve claim

Deviations from the desired total setpoint and deviations from the reserve requirement in the case of a power distribution are evaluated according to the invention with a punishment function which can only assume zero (if fulfilled) or positive function values (if not fulfilled). Together with a cost function that assigns costs to a power distribution depending on the desired total output, this results in a new cost function f for the entire permissible manipulated variable range. The search space for optimization can thus be extended to the entire permissible manipulated variable range; it is not limited to the level of constant power at which the predetermined total setpoint is met.

G. Forecasting the next optimization summation setpoints

In addition to the current summation setpoint, according to the invention, predicted summation setpoints for the next few minutes are used in the form of time profiles or individual values for specific times in order to react to changes in a predictive manner. There is a separate forecast for the change in power consumption (load forecast). Uncontrollable performance Feeds that do not come from the power plants to be managed can also be forecast. The summary results in predicted total setpoints for power plant use optimization (FIG. 9).

H. Forecast of future performance limits

Power plant setpoints that are already fixed before the optimization can be enforced according to the invention by setting the permissible power limits of these power plants to these setpoints. Power generation from power plants that cannot be influenced, e.g. in the arrival or departure are estimated for the next few minutes and interpreted as performance setpoints. The power setpoints and thus also the power limits of the power plants, which are not involved in the optimization of the power plant use, are known in advance. The future power limits of the other power plants can be estimated according to the invention, or they are known from information from the power plants, or they do not change. This provides forecasts for the future performance limits of all power plants, which can then be used in the preview together with the forecast total setpoints.

K. Predictive search space restrictions

It is assumed that predicted total values are given for the next few minutes (Fig. 9). In order to prevent that, by optimizing the current total nominal value, the total values predicted for later are no longer or not optimally reached, the search space of the optimization for the current total setpoint value according to the invention, based on the previously permissible and achievable search space.

A hard search space restriction may be the prohibition of smaller setpoints of the power plants if the predicted total setpoint is greater than the current in a certain time, or the prohibition of larger setpoints of the power plants if the predicted total setpoint is less than the current in a certain time . A soft search space restriction exists if the setpoint changes of the power plants, which are directed against the tendency of the total setpoint, are prohibited depending on the size of the change in the total setpoint. 8 shows an example of the resulting restriction of the search space S ₂ to the search space S ₃ .

L. Construction of a family of tracks in the search area

Predicted total setpoints are given for the next few minutes (FIG. 9). If, based on P ₀ , the next optimal power distribution is determined within one time interval from one to the next forecast value P ₁ , P ₂ , P ₃ , a path results in the search area (FIG. 10). If one or more of the forecast values P ₁ , P ₂ , P _{3 are} omitted in between on this path and the omitted forecast values are determined, the substitute power distribution for which the distance in the search area from the previous to the next power distribution is minimal (shortest route), but at the same time In this way, the current forecast value is met (shortest detour), then different paths are created in the search area. In the event that between the previous, the ancient and the next If the distribution of power is of different lengths, the length of the tracks is also determined according to the invention with the reciprocal of the associated time difference by the sum of the products of the individual lengths of the sections (FIG. 14).

In order to generate a finite number of paths, any possible combination of intermediate forecast values is omitted in succession according to the invention. This creates a family of paths in the search area, which contains both the step-by-step optimized path and the shortest path, taking into account the intermediate total setpoints E ₀ , E ₁ , E ₂ , E ₃ (FIGS. 10 to 13). If an interim total setpoint E _i cannot be reached on a step-by-step optimized path, there are alternative paths according to the invention which come as close as possible to this total setpoint.

According to the invention, those railways are used as alternatives to the railways constructed in this way, in which a compromise between optimal power distribution, power distribution in the shortest detour and complete omission of the forecast value (E _i ) is chosen for the omitted intermediate points. The optimization can therefore take this circumstance into account in advance when determining the next power distribution according to the invention.

The method according to the invention for optimizing the use of power plants works according to the above to a certain extent on the basis of an evaluation function which supplies a: function value for all permissible power distributions depending on the previous power distributors. This evaluation function is made up of the sum of the extended cost functions of all power plants together. In addition, according to the invention, a punishment function is used for non-fulfillment of the sums and reserve claim. A planned temporary non-fulfillment of these demands can be advantageous, for example, if these demands are better met in the future.

According to the invention, the permissible power limits and setpoint gradients of the power plants limit the search space in which all the permissible power setpoints are based on a given power distribution.

The optimization method according to the invention, taking into account the predicted total setpoints based on a given power distribution, determines a future optimal path of the total setpoints within the search space for a specific time interval by means of a path optimization. If only one summation setpoint without further predicted summation setpoints is used in the optimization, the optimization according to the invention only determines the next optimal power distribution.

If a path optimization is carried out, the optimization method according to the invention works as follows: Starting from the given power distribution (as described above), a set of possible paths is generated in the search space for the specific time interval. To determine the support points of these railways, the optimal power distributions, based on a previous power distribution, and alternative power distributions are determined. According to the invention, there are two variants for the optimization of a power distribution based on a previous power distribution: a) First, the optimization on the level of constant power within the search space on which the predetermined total setpoint value is fulfilled determines the power distribution with the smallest value of the evaluation function. On the basis of the point P _{0 found in this way} , the optimization then determines the point P ₁ on the level of constant power within the search space, which has the smallest punishment function value when the evaluation function deteriorates as little as possible compared to the starting point (step-by-step increase in the manipulated variable reserve taking into account the sum requirement) . b) The optimization determines the point P within the search space for which the sum as a valuation function and punishment function is minimal, without first trying to meet only the predefined sum setpoint.

According to the invention, an iterative nutritional method is used to calculate the replacement power distributions for omitted total setpoints by the shortest detour. With a compromise between an optimal power distribution, a power distribution on the. shortest detour and a complete omission of the forecast value without replacement, the optimal and the replacement service allocation for the shortest detour are calculated. Of these bans, the one is selected as the optimal trajectory which has the smallest evaluation and punishment function values in a time-weighted average. The first point of this path is the searched next optimal power distribution. According to the invention, local, global and heuristic optimization methods are suitable for optimizing a power distribution, in particular also a modified step-by-step algorithm (a special step-by-step gradient method).

The method according to the invention is distinguished, among other things, by the following advantages when used to optimize the current use of power plants compared to conventional methods: In contrast to conventional methods, the manipulated variable reserve is kept available in a cost-optimal manner. In particular, a larger manipulated variable reserve can be used than when assigning fixed control bands to selected power plants; this saves costs. At the same time, power plant management is becoming more economical. The restrictions of the search space for the optimization are less restrictive, so that the optimization for the distribution of the total power can determine additional and cheaper power distributions. - The control quality of the power-frequency control improves according to the invention by a time-independent and uniform manipulated variable reserve. In addition, the method according to the invention is characterized by a higher flexibility with changing control expectations. because the reserve requirements can be continuously adjusted to the expected fluctuations in performance in the supply area. - The forward-looking restriction of the search space of the optimization according to the invention or the planning according to the invention of an optimal trajectory of the setpoints in the search space of the optimization avoids undesired flipping effects (opposite setpoint tendencies of individual power plants). In addition, according to the invention, it is ensured in advance that the reserve requirements and the predicted total setpoints are best met in the next few minutes or in the best way on average. If predicted total setpoints cannot be reached, total setpoints are deliberately over or under-fulfilled in advance.

- The setpoint curves of the individual power plants are calmed according to the invention by punishing gradients and gradient changes, so that there is a gentler power plant operation.

The optimization method according to the invention is particularly suitable together with a method for the central control of a power plant park, which is described in the simultaneously filed German patent application "Process for regulating the performance of a power plant park" by the same applicant (file number 40 713 K). The degrees of freedom, parameters and specifications of the procedure can be used by external modules for monitoring, quality assessment or preview, e.g. for a change in the reserve requirements depending on the expected size of the power fluctuations and the control quality achieved for an adaptation of the extended box functions to punish changes too restless be value history. of individual power plants or for an automatic change of permitted power limits and gradients of individual power plants.

For additional superordinate facilities for monitoring, quality assessment or foresight, methods for pattern recognition, event recognition, cluster analysis, neutron networks, fuzzy systems, simulation systems or expert systems are available, in particular a fuzzy quality measure which is linked to the process and which evaluates the process on-line and is used by the process operator or a direct feedback to intervene in the process.

There are particular advantages for the invention when using positive and negative fuzzy rules and when using a parameterized fuzzy inter-frequency filter, in which the parameter specifies whether the rules should be interpreted harder or softer, at the higher level for monitoring and evaluation ( see German patents 43 08 083 and 44 16 465).

The method according to the invention is particularly suitable for optimizing the use of power plants, preferably with a digital computer system, an automation system or a simulation system, parts of the method also being able to be implemented with different systems.

Claims

Claims:

1. Method for sequential feedforward control of a process P which consists of n parallel subprocesses P ₁ , P ₂ , ... P _n with input variables u ₁ , u ₂ , ... u _n and output variables y ₁ , y ₂ , ... y _n exists, a) whereby for the input variables individual

Restrictions on the form u _i , _min ≤ u _i ≤ u _{i, max} and,

aa) from which the distances Sum u _i and ∑ v _i differ from the summary limits ∑ u _{i, min} , ∑ u _{i, max} and Σ v _{i, min} Σ v _{i, max} are the total reserves for the sums ∑ u _i and Σ v _i ,

ab) the output variable y of the process P in the form y = y ₁ + y ₂ + ... y _n arises and

ac) process operation is evaluated by a cost function f, which results from the sum of partial cost functions f _i (u _i ),

wherein to achieve a desired output waveform y _A (t) the sub-processes P _i with input waveforms u _i (t) are applied and these curves u _i (t) of the sum of the ven a pilot control device A and Regulator B generated components u _{i, A} (t) and u _{i, B} (t) are formed.

c) wherein the controller B reacts to a predetermined control variable eg, which arises from disturbances and deviations of the output variable y from the desired output variable profile y _A (t), and

d) the pilot control device _A uses a setpoint curve e _A (t) to be specified, which assigns a setpoint e _A to the current point in time and future points in time, characterized in that.

e) that in the event that the next setpoint e _A for the pilot control A cannot be completely divided into the input variables u _i, A within a cycle time, without adding up with the current input variables u _{i, B} , starting from predefined input variables u ' _i , the next input variables u _i violate the individual restrictions, the maximum possible input variables u _{i, A are} chosen, and

f) that for the other case, for which the

Setpoint value e _A of pilot control A can be divided completely into the input variables u _{i, A} in one or more ways, a division is selected for which there is a total reserve for the regulation of controller B,

g) that the latter division is determined so that within a predetermined range of values

[u _{B, min} ; u _{B, max} ] for the sum of the input variables u _{i, B} , starting from the current sum ∑ u _{i, B} , with constant input variables u _{i, A} while observing the individual restrictions of the input variables u _i

ga) the sum of the input variables u _{i, A} with the minimum rate of change v _{B, min reaches} the value u _{B, min} and with the maximum rate of change v _{B, max} the value u _{B, max} , and

gb) the cost function f for the division u _{i, A is also} minimal.

2. The method according to claim 1, characterized in that only selected input variables u _{i, B are used} for a contribution to the total reserve for control B.

3. The method according to claim 1 or 2, characterized in that the contributions of the input larger. u _{i, B} that to the summary. Reserve for control B to make a contribution, to be limited to individual maximum values u _{i, B, min} , u _{i, B, max} , v _{i, B, min} , and v _{i, B, max} .

4. The method according to any one of claims 1 to 3, characterized in that partial cost functions f _i (u _i ) include or approximate the operating costs of the processes P _i .

5. The method according to any one of claims 1 to 4, characterized in that current or previous input variables u _i , current or previous output variables y _i and / or available current or previous states of the processes P _{i are} integrated into the partial cost functions f _i .

6. The method according to any one of claims 1 to 5, characterized in that both the value of a punishment function k, which assigns a clear function value to the difference between the sum of all input variables u _{i, A} and the target value e _A , and the sum of the partial cost functions f _i can be minimized by choosing that division of the setpoint value e _A into the input variables u _{i, A} for which a new function f, which is a linear combination of the functions f _i and k, becomes minimal.

7. The method according to any one of claims 1 to 6, characterized in that, in addition, the value of a punishment function h, which assigns the difference between a measure for the existing total reserve and a measure for the required total reserve a unique function value, is minimized, in which that distribution of the sell value e _A to the input variables u _{i, A is} chosen for which a new function f, which is a linear combination of the functions f _i and h or f _i , k and h, becomes minimal.

8. The method according to any one of claims 1 to 7, characterized in that depending on the desired future setpoint e _A (t + T) for the pilot control A after a predetermined time T from the possible next divisions u _{i, A} , in which a change in the setpoint value e _A within the time T no change in the input variables u _{i, A} opposite to this change, starting from u ' _{i, A} , occurs, the one with a minimum function value is selected.

9. The method according to any one of claims 1 to 7, characterized in that the setpoint e _{A of} the pilot control A at a fixed number of desired future setpoints for a defined time range T, taking into account future individual restrictions for the input variables u _{i in} such a way on the input variables u _{i, A is} divided in such a way that the function f becomes minimal over time range T by constructing a family of sequences of input variables u _A = (u _{1, A} , u _2A , ... u _{n, A} ) , by all possible combinations either by step-by-step procedure to determine the next division u _A or from the determination of the next division u _A by minimizing the path length between the previous division u _A and the subsequent division u _A while the target value e _{A is} now completely divided and respect the individual. Restrictions or if the target value e _{A is} completely ignored from the determination of the division u _A on the direct connection between the previous division u _A and the subsequent division ur, or from a compromise: these three Possibilities arise from which the first division u _{A of} the sequence of divisions u _{A is} selected, which has the smallest functional value of all constructed divisions on a time-weighted average.

10. The method according to any one of claims 1 to 9, characterized in that the division u _{A is determined} cyclically, acyclically or cyclically with acyclic processing requirements if the division u ' _{A is} the result of the previous cycle or process run.

11. Use of the method according to one of claims 1 to 10 for optimizing the use of a power plant by a power plant management system, characterized in that the function e _A (t) of the load forecast corresponds to the current power plant use optimization, the subprocesses P _{i are} power plants and the device B the Control power distribution of the setpoint from the power-frequency control is.

12. Use of the method according to claim 11, characterized in that in addition superordinate devices for monitoring, quality assessment or foresight the degrees of freedom, parameters, specifications or strategies of the optimization method are used or changed.

13. Power plant management system, which is optimized by the method according to one of claims 1 to 12, characterized in that it is equipped with a digital computer system, an automation system and / or a simulation system.

14. Power plant management system according to claim 13, characterized in that parts of the method are equipped with different or parallel systems.

15. Power plant management system according to claim 13 or 14, characterized in that methods of pattern recognition, event recognition, cluster analysis, neural networks, fuzzy systems, simulation systems or expert systems serve as additional superordinate devices.

16. Power plant management system according to claim 15, characterized in that the additional superordinate devices contain fuzzy controllers with positive and negative rules or parameterized inference filters.

17. The method, in particular according to one of claims 1 to 10, for optimizing the current power plant use by a power plant management system with an overall rule rule applicable to the guided power plants, which limits the control power sum up and down, and with individual gradient reserves for each guided power plant the positive and negative gradients of the power setpoints, characterized in that the current distribution of the setpoints and / or the actual power is used as the starting point for determining the optimal power distribution for the next time step, that within the summary rule band independent of the individual gradient reserves of the individual Power plant a minimum control speed for the time until the Errei Chen the control band limit is set so that each power plant involved in the control is assigned a maximum permissible power component for the control, and that the summation setpoint curves of the services and the associated summation setpoint gradients are restricted for the individual power plants.

18. The method according to claim 17, characterized in that a look ahead to changes in the totals of the setpoint changes (total setpoint) in the form of time profiles or of individual values for certain times predicted total setpoints - for example for the next few minutes - is reacted.

19. The method according to claim 17 or 18, characterized in that before the optimization fixed power station setpoints are forced by the fact that the permissible power limits are set to these setpoints.

20. The method according to any one of claims 17 to 19, characterized in that the area (search space) of the optimization for the respective current total setpoint, based on the previously permissible and achievable search space., Is restricted in advance.

21. The method according to claim 2, characterized in that larger soli values of the power plants are prohibited if the predicted SumoIIwert m is less than the current time.

22. The method according to claim 20, characterized in that the setpoint changes of the power plants, which are directed against the tendency of the total setpoint, are prohibited depending on the size of the total setpoint change.