초록 ▼

The invention concerns a method for the computerized control and/or regulation of a technical system (T). Within the context of the method according to the invention, there is implemented in a preset regulating process (CO1, CO2) an exploration rule (EP) by means of which new, as yet unknown, states

The invention concerns a method for the computerized control and/or regulation of a technical system (T). Within the context of the method according to the invention, there is implemented in a preset regulating process (CO1, CO2) an exploration rule (EP) by means of which new, as yet unknown, states (x) of the technical system (T) are started, a simulation model (SM) of the technical system (T) checking whether the actions (a2) of the exploration rule (EP) lead to sequential states (x′) lying within predetermined thresholds. Only in that case is the corresponding action (a2) performed according to the exploration rule (EP) on the technical system. The method according to the invention enables new states to be explored within the framework of the operation of a technical system, it being ensured through checking of appropriate thresholds that the exploration is carried out imperceptibly and does not lead to incorrect operation of the technical system. The method according to the invention is suitable in particular for controlling or regulating gas turbines or wind turbines.

대표청구항 ▼

1. A method for computer-assisted open-loop control, closed-loop control, or open-loop and closed-loop control of a technical system, wherein the open-loop control, the closed-loop control, or the open-loop and closed-loop control carries out, at respective times, actions comprising one or more acti

1. A method for computer-assisted open-loop control, closed-loop control, or open-loop and closed-loop control of a technical system, wherein the open-loop control, the closed-loop control, or the open-loop and closed-loop control carries out, at respective times, actions comprising one or more action variables at the technical system as a function of one or more states of the technical system, comprising in each case one or more state variables, the method comprising: determining, by a predefined regulating method, at respective times, first actions that are to be carried out on the technical system as a function of at least the state of the technical system at the respective time;determining, by at least one exploration rule, one or more second actions to be carried out on the technical system as a function of at least the state of the technical system at the respective time, the at least one exploration rule configured to determine the one or more second actions to be carried out to explore states that have not been adopted by the technical system;predicting a set of resulting subsequent states of the technical system based on a simulation model of the technical system for each second action;checking, for each set of subsequent states, whether the respective set of subsequent states satisfies a predefined threshold value criterion that defines thresholds to be complied with for at least one portion of the state variables of the respective set of subsequent states; andsupplying a second action to the predefined regulating method when one or more sets of subsequent states satisfy the threshold value criterion, wherein for the second action, the predicted set of subsequent states satisfying the threshold value criterion,wherein after supplying the second action, the predefined regulating method is configured to carry out the supplied second action instead of the first action, or to carry out the first action in combination with the second action, the technical system,wherein in an event of none of the sets of subsequent states satisfy the threshold value criterion, the predefined regulating method carries out the first action, one or more modified second actions are determined, or the predefined regulating method carries out the first action and one or more modified second actions are determined,wherein, for the one or more modified second actions, the method further comprises: predicting a set of resulting modified subsequent states of the technical system based on a simulation model of the technical system for each modified second action;checking, for each set of modified subsequent states, whether the respective set of modified subsequent states satisfies a predefined threshold value criterion that defines thresholds to be complied with for at least one portion of the state variables of the respective set of modified subsequent states;supplying a modified second action to the predefined regulating method when one or more sets of modified subsequent states satisfy the threshold value criterion, wherein for the modified second action, the predicted set of modified subsequent states satisfying the threshold value criterion,wherein after supplying the modified second action, the predefined regulating method carries out the supplied modified second action instead of the first action, or carries out the first action in combination with the modified second action, at the technical system. 2. The method of claim 1, wherein each of the one or more determined second actions constitutes a variation of the first action, and wherein the supplied second action is carried out at the technical system. 3. The method of claim 1, wherein the threshold value criterion specifies, for at least a portion of the state variables of the subsequent states, respective value ranges within which values of the portion of the state variables must lie. 4. The method of claim 3, wherein the value ranges depend on the values of the state variables of the subsequent states and define a percentage deviation from the values. 5. The method of claim 1, wherein the at least one exploration rule comprises a rule that randomly selects the second action, a rule that is based on expert knowledge, an ε greedy rule, or any combination thereof. 6. The method of claim 1, wherein the simulation model is learnt with a machine learning method and is based on a learnt neural network. 7. The method of claim 1, wherein the predefined regulating method is a table-based regulating method in which respective states of the technical system are assigned to first actions by a table, an action selection rule that is learnt with a machine learning method, or a combination of a table-based regulating method and an action selection rule that is learnt with a machine learning method, and wherein the action selection rule specifies a variation of an action that is determined with the table-based regulating method, and the first action is that action of the table-based regulating method that is varied by the variation. 8. The method of claim 1, wherein the technical system is a gas turbine, a wind turbine, or the gas turbine and the wind turbine. 9. The method of claim 8, wherein the technical system comprises the gas turbine or the gas turbine and the wind turbine, and wherein a state of the gas turbine comprises one or more state variables, the one or more state variables comprising a temperature, a pressure, or the temperature and the pressure at one or more locations on the gas turbine, an air humidity at one or more locations on the gas turbine, combustion chamber accelerations in the gas turbine, emission of pollutants by the gas turbine, power generated by the gas turbine, or any combination thereof. 10. The method of claim 8, wherein the one or more action variables of the first actions, the second actions, or the first actions and the second actions to be carried out at the gas turbine comprise a change in a setting of one or more fuel injection valves, a change in a position of one or more vanes, a change in one or more cooling air streams in the gas turbine, or any combination thereof. 11. The method of claim 9, wherein the portion of state variables in the checking, for which thresholds that are to be complied with are defined, comprise the combustion chamber accelerations in the gas turbine, the emission of pollutants by the gas turbine, or a combination thereof. 12. The method of claim 8, wherein the technical system comprises the wind turbine or the gas turbine and the wind turbine, and wherein a state of the wind turbine comprises one or more state variables, the one or more state variables comprising a temperature, a pressure, an air humidity, or any combination thereof at one or more locations on the wind turbine, a strength of wind at the wind turbine, an amplitude of oscillations of a tower of the wind turbine, mechanical loading of rotor blades of the wind turbine, power generated by the wind turbine, or any combination thereof. 13. The method of claim 8, wherein the technical system comprises the wind turbine or the gas turbine and the wind turbine, and wherein the one or more action variables of the first actions, the second actions, or the first actions and the second actions to be carried out at the wind turbine comprise a change in angles of attack of rotor blades of a rotor of the wind turbine, a change in orientation of the rotor of the wind turbine with respect to the wind, or a combination thereof. 14. The method of claim 12, wherein the at least one portion of state variables in the checking, for which thresholds that are to be complied with are defined, comprise the amplitude of the oscillations of the tower of the wind turbine, the mechanical loading of the rotor blades of the wind turbine, or a combination thereof. 15. The method of claim 6, wherein the learnt neural network is a recurrent neural network. 16. The method of claim 9, wherein the temperature, the pressure, or the temperature and the pressure at one or more locations on the gas turbine comprises an ambient temperature, an ambient pressure, a compressor temperature, a compressor pressure, a temperature in a combustion chamber, a pressure in the combustion chamber, or any combination thereof. 17. The method of claim 8, wherein the emission of pollutants by the gas turbine comprises emission of nitrogen oxide. 18. The method of claim 12, wherein the temperature, the pressure, the air humidity, or any combination thereof at the one or more locations on the wind turbine comprises an ambient temperature, an ambient pressure, an ambient air humidity, or any combination thereof. 19. A computer program product having a program with instructions that are stored on a non-transitory machine-readable carrier for computer-assisted open-loop control, closed-loop control, or open-loop and closed-loop control of a technical system, when the program runs on a computer, wherein the open-loop control, the closed-loop control, or the open-loop and closed-loop control carries out, at respective times, actions comprising one or more action variables at the technical system as a function of one or more states of the technical system, comprising in each case one or more state variables, the instructions comprising: determining, by a predefined regulating method, at respective times, first actions that are to be carried out on the technical system as a function of at least the state of the technical system at the respective time;determining, by at least one exploration rule, one or more second actions to be carried out on the technical system as a function of at least the state of the technical system at the respective time, the at least one exploration rule configured to determine the one or more second actions to be carried out to explore states that have not been adopted by the technical system;predicting a set of resulting subsequent states of the technical system based on a simulation model of the technical system for each second action; checking, for each set of subsequent states, whether the respective set of subsequent states satisfies a predefined threshold value criterion that defines thresholds to be complied with for at least one portion of the state variables of the respective set of subsequent states; and supplying a second action to the predefined regulating method when one or more sets of subsequent states satisfy the threshold value criterion, wherein for the second action, the predicted set of subsequent states satisfying the threshold value criterion, wherein after supplying the second action, the predefined regulating method is configured to carry out the supplied second action instead of the first action, or to carry out the first action in combination with the second action, at the technical system, wherein in an event of none of the sets of subsequent states satisfy the threshold value criterion, the predefined regulating method carries out the first action, one or more modified second actions are determined, or the predefined regulating method carries out the first action and one or more modified second actions are determined, wherein, for the one or more modified second actions, the method further comprises: predicting a set of resulting modified subsequent states of the technical system based on a simulation model of the technical system for each modified second action;checking, for each set of modified subsequent states, whether the respective set of modified subsequent states satisfies a predefined threshold value criterion that defines thresholds to be complied with for at least one portion of the state variables of the respective set of modified subsequent states;supplying a modified second action to the predefined regulating method when one or more sets of modified subsequent states satisfy the threshold value criterion, wherein for the modified second action, the predicted set of modified subsequent states satisfying the threshold value criterion, wherein after supplying the modified second action, the predefined regulating method carries out the supplied modified second action instead of the first action, or carries out the first action in combination with the modified second action, at the technical system.