10, in particular to quickly regulate rapid speed fluctuations.

FIG. 3 shows the plurality of closed-loop controllers 101 of the closed-loop speed control module 10. In the example of FIG. 3, specifically 3 (or, in general, n) closed-loop controllers 101 are shown, but a larger number of closed-loop controllers 101 is also conceivable, (e.g., more than 10 closed-loop controllers). A cascade of n closed-loop controllers may thus be used. Each controller of the closed-loop controllers 101, (two of which are also shown somewhat larger in FIG. 4), includes a P element, an I element, and a D element. However, it would also be conceivable for one or more of the closed-loop controllers to include only one P element, only one I element, or only one D element, or any two such elements. The PID closed-loop controller may also be configured structurally, for example, by an anti-wind-up device that may be used to achieve a numerical limitation of the closed-loop controller output and thus a more dynamic closed-loop controller structure. The influence on fluctuating closed-loop control differences may be reduced by limiting the output signal of the D component. Each closed-loop controller 101 receives a difference between the speed signal (e.g., the current speed of the turbine 205) and the setpoint speed signal (e.g., the setpoint value of the speed of the turbine 205) at a respective closed-loop controller input RE. The difference is calculated by a subtractor 106. Based on this difference, the P element, the I element, and the D element each provide an individual manipulated value. These individual manipulated values are added to a manipulated value and output as a manipulated value signal at a closed-loop controller output RA of the respective closed-loop controller 101. This is done in parallel for all closed-loop controllers 101.

A weighting block 103 is assigned to each closed-loop controller 101. The speed signal is supplied to each of the weighting blocks 103. Each weighting block 103 stores a weighting function, which assigns a weighting to each speed. For example, the weighting is between 0 and 1. If a speed signal is supplied to the weighting block 103, then this outputs a weighting signal that indicates the weighting corresponding to the speed of the speed signal. Any continuous function may be used as a weighting function. A triangle, rectangle, Hanning, Hamming, or Blackman window function or a Gaussian distribution function is expediently used.

The weighting of the weighting signal is multiplied by the manipulated value of the manipulated value signal in a multiplier 105. This is done in parallel for all closed-loop controllers 101. The weighted manipulated values of the individual closed-loop controllers 101 are added to the output signal in an adder 104, wherein the signal is then output at the output 102.

As explained in more detail below, the plurality of closed-loop controllers is set differently in comparison to one another. The different closed-loop controllers 101 thus provide different manipulated values for the same entered speed difference from one other. Specifically, at least one respective P component, I component, and/or D component of the plurality of closed-loop controllers 101 is set differently.

FIG. 4 shows, for example, speed-dependent weighting values h(n) in a range of speeds n from an idle speed n_Idlevia an exemplary setpoint speed n_Sollup to a maximum speed n_Max. Several weighting functions are shown here, each of which is designed by way of example in the form of a Hamming window function. Each weighting function thus has a range in which it makes a contribution, and in the other ranges of the speed is equal to 0 or essentially equal to 0. Each of the weighting functions, which in the example of FIG. 4 are formed equally and arranged equidistantly with one another, is stored in a weighting block 103. Thus, the individual closed-loop controllers 101 are only activated in certain speed ranges. The individual closed-loop controllers 101 are therefore assigned to different speed ranges.

In the specific design according to FIG. 4, the weighting functions are evenly distributed across the entire speed range. However, it is also conceivable that certain (in particular predetermined) ranges, (e.g., the setpoint speed range), are resolved more finely, that is to say the weighting functions may be compressed or stretched in terms of width. The total result of the weighting functions is not more than 1 in this case.

A PI controller 101 may be used at the idle speed n_Idle, at which the D component is deactivated. Simple P controllers 101 may be used in the speed range between n_Idleand n_Soll, in which the D and I components are deactivated. The P components may vary for the individual closed-loop controllers 101. A PI controller or a PID controller 101 may be used for the setpoint speed n_Soll.

The top graph of FIG. 5, corresponding to FIG. 4, shows weighting functions which in this case are partly different from one another, however. For example, the weighting functions in the range of the setpoint speed n_Sollthus have a smaller width than in areas located further away. This enables a finer adjustment in the range of the setpoint speed to be achieved.

Furthermore, a bottom graph shows the gain K_pof the P element P of the respectively associated closed-loop controller 101. As may be seen, the gains K_pof the individual closed-loop controllers 101 are set differently and have different magnitudes. In the present case, for example, an increased gain K_pis provided in a medium speed range in order to be able to pass through a resonance range quickly when starting up the turbogenerator 2. In addition, increased gains K_pare provided on both sides of the setpoint speed in order to keep the setpoint speed precise.

Different types of weighting functions are conceivable, (e.g., also triangle windows, Hanning functions, rectangular functions, or generally window functions). At each speed, the sum of the weighting functions results in 1. Optionally, if the maximum speed is exceeded, the output signal may be set to 0 in order to enable fast downward adjustment.

In this way, adapted closed-loop control parameters may be provided for each operating point. This enables improved closed-loop control quality to be achieved. In addition, linearization may be achieved at the respective operating point. Furthermore, for example, very large, manipulated variables are made possible in lower speed ranges and more moderate manipulated variables in upper speed ranges in order on the one hand to quickly adjust to a target speed and, on the other hand, to achieve a high control quality.

FIG. 6 illustrates the profile of a start-up of the turbogenerator 2 to a desired speed against the time t. In this case, a speed request is illustrated by a thin line. This request is followed by the controlled speed n. A dashed line here illustrates an exemplary path with the closed-loop control described herein in comparison with conventional closed-loop control (solid line).