Patent Application
20070250216
The invention is explained in greater detail below by way of an exemplary embodiment illustrated in the drawings, in which:
Referring now in greater detail to the drawings, an installation 1, for example a coal, oil or gas-fired power-generating plant, a waste incinerator or a cement plant, comprises a furnace 3, which more generally should also be understood to mean a grate, at least one observation device 5, which can image the interior of the furnace 3 (or the grate), preferably other sensors 7, at least one adjusting device 9, and a computer 11. The observation device(s) 5, further sensors 7 and adjusting device(s) 9 are connected to the computer 11.
Fuel, or another material to be converted, is supplied to the furnace 3 along with primary air (or primary oxygen) and secondary air (or secondary oxygen). For the sake of brevity, the fuel, or another material to be converted, (e.g. coal, oil, gas, waste material, lime, or similar material) may be generally referred to as material G. Likewise for the sake of brevity, the primary air (or primary oxygen) and secondary air (or secondary oxygen) may be generally referred to as air L. The supply of the material G and air L is regulated by the adjusting devices 9 which are controlled by the computer 11. A combustion process takes place in the furnace 3. The flame body F that is produced as a result (also any possible emissions from the walls of the furnace 3) is constantly recorded by the observation devices 5. Each of the observation devices 5 comprises an optical access passing through the wall of the furnace 3, a camera or similar device that operates in the optical range or in adjacent ranges of electromagnetic radiation, and it may also in include, for example, a lance or device as disclosed in EP 1 621 813 A1 and/or US 2006/0024628 A1. The entire disclosure of each of EP 1 621 813 A1 and US 2006/0024628 A1 is incorporated herein by reference. Preference is given to a camera having high temporal, local and spectral resolution, such as the camera described, for example, in WO 02/070953 A1 and/or EP 1 364 164 B1. The entire disclosure of each of WO 02/070953 A1 and EP 1 364 164 B1 is incorporated herein by reference.
The images of the flame body F (and of any possible emissions from the walls of the furnace 3) are evaluated in the computer 11, for example using an eigenvalue procedure as described in WO 2004/018940 A1 and/or US 2005/0147288 A1. The entire disclosure of each of WO 2004/018940 A1 and US 2005/0147288 A1 is incorporated herein by reference. EP 1 524 470 A1 describes a process by way of which a few characteristic values can be obtained from a spectrum. The entire disclosure of EP 1 524 470 A1 is incorporated herein by reference. The data obtained from the images of the flame body F, as well as the data from the other sensors 7, which measure, for example, the supply of material G and of air L, concentrations of pollutants in the waste gases, or the concentration of free lime (FCAO), are treated as state variables s(t) that describe (in a time-dependent manner) the state of the system in the installation 1 in general, and the state of the combustion process in particular, and are to be considered as a vector.
A control loop (e.g., system) is defined by the furnace 3 as a (controlled) system, the observation device(s) 5 and the other sensors 7, the computer 11 and the adjusting devices 9. It is also possible to provide a conventional control loop, with just a furnace 3, sensors 7, computer 11 and adjusting devices 9 and without the observation device(s) 5, in which the control function takes account of only a few state variables st (i.e. it is low-dimensional) and is then optimized by including the observation device(s) 5. For example, the system in installation 1 can be regulated to achieve certain setpoints or to achieve a stable process (i.e. smooth, quasi-stationary operation of the installation 1). In both cases, the state described by the actual values of the state variables s(t) is evaluated and, if necessary, suitable adjustment actions (setting actions) are selected which are to be carried out by the adjusting devices 9. For the sake of brevity, the suitable adjustment actions (setting actions), which are selected and are to be carried out by the adjusting devices 9, are referred to as actions ai. In addition to supplying the material G and air L, other activities performed by the adjusting devices 9, and possibly also the taking of a sample, may constitute an action ai within the meaning of the exemplary embodiment of the present invention. Disturbances may also be treated as unintended actions ai. Adjustable combinations of the two above-mentioned control situations are conceivable, which then represent compromises.
The evaluation of the state and the selection of suitable actions ai may, for example, be accomplished by way of a procedure such as that described in WO 02/077527 A1 and/or U.S. Pat. No. 7,035,717. The entire disclosure of each of WO 02/077527 A1 and U.S. Pat. No. 7,035,717 is incorporated herein by reference. At least one neuronal network is implemented in the computer 11, with this network storing as a process model the reactions of the system states to actions ai, i.e. the (non-linear) links between the values of the state variables s(t) at a time t=to and the actions ai which are then taken, on the one hand, and the resulting values of the state variables s(t) at a later (i.e. later by a certain time interval) point in time t=t1(or t1, t2, t3 . . . ), on the other hand, i.e. at as many times t as possible in the past. In this sense, disturbances may also be included in the process model as (unintended) actions ai. An evaluation of the situation, designed as a type of simplified quality, that is independent of the process model, i.e. of the stored links, evaluates the values of the state variables s(t) at a certain point in time t with respect to predetermined optimization targets ri, i.e. to determine how close the system state is to the optimal state at time t. By evaluating a state predicted—by way of the process model as a function of a specific action ai—at a future point in time, it is possible to determine the suitability of the specific action ai for approaching the optimization target ri.
Preferably three (or four) process models are stored (each in their own neuronal network) in the computer 11, with each of the process models containing links learned for one short (t1-to) time interval, for one (or two) medium (t2-to) time intervals, and for one long (t3-to) time interval. Correspondingly, it is thus possible to make short-term, medium-term and long-term predictions. Depending on the installation 1, the time intervals (e.g., t1-to, t2-to, and t3-to) range from a few seconds to several hours. The state variables s(t) should and can usually vary within certain limits, i.e. within an interval, for example between a lower limit value s1 and an upper limit value Sh, around an optimal setpoint so. The values S1, Sh and so can be time-dependent. The short-term, medium-term and long-term predictions serve to estimate the difference between s(t) and the optimal setpoint so (the optimization target ri in the present case would be, for example, for s(t)−so to be equal to 0 or at least to become minimal) and also to determine whether these limits (limit values s1, sh) have been adhered to, as well as to recognize the probable need for actions ai. The temporal development of a state variable s(t) up to time t=to as well as the short-term prediction for t=t1, the medium-term prediction for t=t2 and the long-term prediction for t=t3 are depicted in simplified form in
In order to improve the accuracy, not only are the process models constantly updated by the actual developments of the state variables s(t) as a reaction to actions ai, but also a competition takes place between several process models regarding the quality of the predictions. For this purpose, alternative process models, for example with other topologies, are set up and trained in the background and their predictions compared with the currently used process models in order, if necessary, to replace the currently used process models, in the manner as described, for example, in EP 1 396 770 A1 and/or US 2005/0137995 A1. The entire disclosure of each of EP 1 396 770 A1 and US 2005/0137995 A1 is incorporated herein by reference.
According to the exemplary embodiment of the present invention, it is possible to switch from normal control mode (i.e., so-called setpoint control mode) to disturbance control mode (and back again). In disturbance control mode, the computer 11 sends out test signals so that—without regard for the optimization targets ri—various actions ai are taken in order, in a targeted manner, to approach in various directions initially adjacent states (i.e. adjacent to the respectively current state with regard to the state variables s(t)) and preferably—by successively sequencing the approach—also to reach more distant states. However, in order not to impede, let alone disrupt, the operation of the installation 1, only states within the limits (limit values s1, Sh) of the state variables s(t) are selected as the target, i.e. only actions are selected in response to which the state variables s(t) will probably remain within their limits.
The computer 11 starts “ordinary” disturbance control mode at regular intervals (e.g., approximately every seven days, but at the latest every four weeks). During the ordinary disturbance control mode, as many states as possible are approached, with these states preferably being distributed as uniformly as possible within the limits (e.g., the states are substantially uniformly distributed within the predetermined limits). If the same problems occur frequently (e.g., there is a frequent reoccurrence of one or more problems) during the control procedure (e.g., during the setpoint control mode), the computer 11 starts “extraordinary” disturbance control mode. Such problems exist, for example, when the state variables s(t) frequently tend towards a limit (limit values s1, sh), i.e. the mean value drifts and/or frequently actions ai are needed to compensate for deviations, and/or other inconsistencies occur in the regulation to achieve setpoints (optimization targets ri) and a stable process. In the case of extraordinary disturbance control, it is possible in particular to approach states which are matched to the triggering problems; for example, depending on the solution strategy, the states are selected either oriented towards the problems or in the exactly opposite direction.
In the drawings, for example, a case is depicted where s(t) fluctuates constantly above the optimal setpoint so (
In accordance with the exemplary embodiment of the present invention and as should be apparent to one of ordinary skill in view of the foregoing, the computer 11 (which includes appropriate input and output devices) may control the operation of the installation 1 by virtue of receiving data from and/or providing data (e.g., instructions) to respective components. For this purpose and in accordance with the exemplary embodiment of the present invention, the computer 11 includes or is otherwise associated with one or more computer-readable mediums (e.g., volatile memory and/or nonvolatile memory and/or one or more other storage devices such as, but not limited to, tapes and hard disks such as floppy disks and compact disks) having computer-executable instructions (e.g., one or more software modules or the like), with the computer handling (e.g., processing) the data in the manner indicated by the computer-executable instructions. Accordingly, the computer 11 can be characterized as being schematically illustrative of the computer-readable mediums, computer-executable instructions and other features of methods and systems of the exemplary embodiment of the present invention.
It will be understood by those skilled in the art that while the present invention has been discussed above with reference to an exemplary embodiment, various additions, modifications and changes can be made thereto without departing from the spirit and scope of the invention as set forth in the following claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 06 008 487.8 | Apr 2006 | EP | regional |
