DE102023135427B4 - Method for operating a power device, control device for carrying out such a method and power arrangement with such a control device - Google Patents

Abstract

Method for operating a power device, wherein
- the power device is operated in a predetermined time interval ([t _A , t _E ]) with a transient operating pattern, wherein
- a plurality of chronologically successive control values (X _{[t A ,t E ]} ) at least one control variable $\left(\overrightarrow{x}\right)$ of the power device in the predetermined time interval ([t _A , t _E ]), where
- by means of at least one sensor (11) a plurality of temporally successive measured values (Y _{[t A ,t E ]} ) at least one measured value $\overrightarrow{y}$ is detected in the predetermined time interval ([t _A , t _E ]), where
- the majority of measured values (Y _{[t A ,t E ]} ) are manipulated by means of a power device model (GP) taking into account a dynamic behavior of the at least one sensor (11), whereby a plurality of temporally successive reconstruction measured values $\left(Y^r{}_{\left[t_A,t_E\right]}\right)$ ) of at least one measured value $\overrightarrow{y}$ is obtained, where
- the power device model (GP) based on the majority of reconstruction measurements $\left(Y^r{}_{\left[t_A,t_E\right]}\right)$ and the majority of control values (X _{[t A ,t E ]} ) is adjusted, whereby
- the power device is operated based on the adapted power device model (GP).

Description

The invention relates to a method for operating a power device, a control device for carrying out such a method and a power arrangement with a power device and such a control device.

It is known that an adaptive model-predictive control of the internal combustion engine's manipulated variables is used to operate an internal combustion engine. Hyperspace control uses models of the internal combustion engine's combustion variables to determine the manipulated variables. The models are typically determined and/or adapted based on steady-state operation of the internal combustion engine, particularly during test bench experiments. The disadvantage is that these models cannot be adapted during transient operation of the internal combustion engine. Furthermore, these models are therefore not suitable for correctly determining the combustion variables during transient operation of the internal combustion engine and thus for optimally adjusting the manipulated variables.

From the German patent specification DE 10 2020 125 533 B3 A device is known that has a signal interface, a processor unit, and a memory unit. The signal interface is configured to receive a measurement signal representing first sensor data and second sensor data of a power transformer. Furthermore, the processor unit is configured to determine a value for a temperature in the power transformer that is not detected by sensors based on the sensor data and a simulation model for the power transformer.

The invention is therefore based on the object of providing a method for operating a power device, a control device for carrying out such a method and a power arrangement with a power device and such a control device, wherein the disadvantages mentioned are at least reduced, preferably do not occur.

The object is achieved by providing the present technical teaching, in particular the teaching of the independent claims as well as the preferred embodiments disclosed in the dependent claims and the description.

The problem is solved by providing a method for operating a power device. The power device is operated in a predetermined time interval with a transient operating pattern, wherein a plurality of temporally successive control values of at least one control variable of the power device are specified in the predetermined time interval. A plurality of temporally successive measured values of at least one measured variable are acquired in the predetermined time interval by means of at least one sensor. Furthermore, the plurality of measured values are manipulated using a power device model, taking into account the dynamic behavior of the at least one sensor, such that a plurality of temporally successive reconstruction measured values are obtained. The power device model is then adapted based on the plurality of reconstruction measured values and the plurality of control values, wherein the power device is operated based on the adapted power device model. Due to the dynamic behavior of the at least one sensor, the measured values acquired at a specific time do not reflect the actual values of the at least one measured variable at the location of the at least one sensor at the specific time. Therefore, the control value and the measured value at the specific time cannot be used as data points for adapting the power device model. Advantageously, the dynamic behavior of the at least one sensor is taken into account in the reconstruction measured values, so that the reconstruction measured values of the specific time and the control values of the specific time can be used as data points for adapting the power device model. Thus, the power device model is advantageously adapted during transient operation of the power device. In addition, optimal control values of the at least one control variable can also be determined for transient operation of the power device based on the adapted power device model.

In the context of the present technical teaching, a power device is understood to mean, in particular, a device that is configured to provide power, in particular electrical and/or mechanical power, or to convert or consume power. The power device can thus be designed as a power supply device or as a power conversion device. A power supply device is preferably understood to mean a device that provides power using electrical, mechanical, chemical or electrochemical energy - or another form of energy - such as electrical and/or mechanical power.

A power conversion device is preferably understood to be a device that uses or consumes power, for example electrical or mechanical power, to convert or store energy, for example to provide chemical energy in the form of certain substances such as hydrogen or methanol, or electrochemical energy using electrical energy. The power device can be an internal combustion engine, a combined internal combustion engine-generator device, i.e. a genset, a fuel cell, an energy storage device, for example a battery, or an electrolyzer. However, the power device can also be a larger, complex system, for example comprising a plurality of the aforementioned devices, or in particular also a data center or a microgrid. The power device can also be a controllable or regulatable load on an electrical network.

In one embodiment, the predetermined time interval is a period from the initial time to an end time following the initial time, where t _A denotes the initial time and t _E denotes the end time.

In the context of the present technical teaching, a manipulated variable is preferably understood to mean a variable that is suitable for controlling an actuator of the power device. Furthermore, a manipulated variable and/or a plurality of manipulated variables is denoted below by the symbol $\overrightarrow{x}$ The following applies: $\overrightarrow{x}=x_1$ for a single control variable and $\overrightarrow{x}={\left[\begin{array}{ccc}{x}_1& \dots & x_i\end{array}\right]}^T$ $$ for a plurality of manipulated variables with a first manipulated variable x ₁ and an i-th manipulated variable x _i , where i ≥ 2. It is possible for the manipulated variable to be directly suitable for controlling the actuator, for example because it is given as a specific voltage or current that can be directly switched to the actuator in order to control it; alternatively, it is possible for the manipulated variable to be suitable for deriving, in particular calculating, at least one further variable for directly controlling the actuator. For example, the manipulated variable can be a fuel mass to be introduced into a combustion chamber of an internal combustion engine, which can be converted into at least one control variable for controlling an injector.

The at least one actuator can be an actuator or actuator of the power device. In particular, the actuator can be an actuator of an engine block of an internal combustion engine, for example, an injector, a valve, or a flap. However, the actuator can also be an actuator outside the power device, in particular outside an engine block, for example, an actuator provided for influencing an externally provided cooling circuit, for example, a valve, a pump, or the like, or an actuator of a transmission or an electrical device to which the power device is operatively connected.

In the context of the present technical teaching, a control value is preferably understood to be a value of the control variable or of the plurality of control variables that is set and/or specified within a specific period of time or at a specific point in time. Furthermore, a control value is denoted below by the symbol $\overrightarrow{x}\left(t\right)$ where t denotes the specific period of time or the specific point in time. Furthermore, regardless of the number of manipulated variables, the manipulated value—especially represented in vector form—is referred to as the manipulated value.

In the predetermined time interval [t _A , t _E ], at a plurality of times t _n with t _A ≤ t _n ≤ t _E a control value $\overrightarrow{x}\left(t_n\right)$ Therefore, the majority of control values in the predetermined time interval are stored in a matrix of the form $$X_{\left[t_A,t_E\right]}={\left(\overrightarrow{x}\left(t_n\right)\right)}_{n=\mathrm{1,}\dots ,k}$$ shown.

In the context of the present technical teaching, a measured variable is preferably understood to mean a variable that can be detected and/or measured by means of a sensor. Furthermore, a measured variable and/or a plurality of measured variables is denoted below by the symbol $\overrightarrow{y}$ The following applies: $\overrightarrow{y}=y_1$ for a single measurement and $\overrightarrow{y}={\left[\begin{array}{ccc}{y}_1& \dots & y_j\end{array}\right]}^T$ $$ for a plurality of measured variables with a first measured variable y ₁ and a j-th measured variable y _j , where j ≥ 2.

In the context of this technical teaching, a measured value is preferably understood as a value of the measured quantity recorded at a specific point in time. Furthermore, a measured value is referred to below with the symbol $\overrightarrow{y}\left(t\right)$ where at time t, which marks the specific period or the specific time of the control value, the measured value is recorded, for example, in a control device. Furthermore, the measured value—in particular, represented in vector form—is referred to independently of the number of measured variables.

In the predetermined time interval [t _A , t _E ], at the plurality of times t _n with t _A ≤ t _n ≤ t _E a measured value $\overrightarrow{y}\left(t_n\right)$ Therefore, the majority of measured values in the predetermined time interval are stored in a matrix of the form $$Y_{\left[t_A,t_E\right]}={\left(\overrightarrow{y}\left(t_n\right)\right)}_{n=\mathrm{1,}\dots ,k}$$ shown.

In the context of the present technical teaching, a reconstruction measured value is preferably understood to be a measured value manipulated taking into account the dynamic behavior of the at least one sensor. Furthermore, a reconstruction measured value is denoted below by the symbol ${\overrightarrow{y}}^r\left(t\right)$ where t indicates the time point that characterizes the specific period or the specific time point of the control value. Furthermore, regardless of the number of measured variables, the term "reconstruction measured value" is used—especially when represented in vector form.

In the predetermined time interval [t _A , t _E ], for the majority of times t _n with t _A ≤ t _n ≤ t _E a reconstruction measurement value ${\overrightarrow{y}}^r\left(t_n\right)$ Therefore, the majority of reconstruction measurement values in the predetermined time interval are stored in a matrix of the form $$Y_{\left[t_A,t_E\right]}^r={\left({\overrightarrow{y}}^r\left(t_n\right)\right)}_{n=\mathrm{1,}\dots ,k}$$ shown.

In the context of the present technical teaching, a vector is preferably understood to mean a combination of at least one variable and/or at least one value, regardless of how the at least one variable and/or at least one value is represented. In principle, any number or combination of variables and/or values can be understood or represented as a vector with at least one component, where the number of vector components corresponds to the number of variables and/or values.

In one embodiment, the power device model comprises a nominal model and a detailed model. Alternatively, the power device model consists of the nominal model and the detailed model.

Preferably, the detailed model of the power device model is adapted based on the plurality of reconstruction measured values and the plurality of control values.

In one embodiment, a Gaussian process model is used as the power device model. Gaussian process models are particularly suitable for controlling a power device: Compared to polynomial-based models, they are easier to adapt to new or changed data points in the application field, and they exhibit more suitable and physically correct behavior in the boundary regions of the given parameter space. Compared to physical models, they require significantly less computational effort. Furthermore, they enable the direct use of test bench data. In one embodiment, such a Gaussian process model is given by stored data points (X _b , Y _b ), obtained, for example, in test bench tests, where X _b ∈ ℝ ^n×m are n input variables for m different operating states and Y _b ∈ ℝ ^m×k are k output variables for the m different Operating states are specified. The control values are preferred $\overrightarrow{x}\left(t\right)$ - the predetermined time interval and/or other times - than the input variables X _b and the associated measured values $\overrightarrow{y}\left(t\right)$ and/or the associated reconstruction measurements ${\overrightarrow{y}}^r\left(t\right)$ as output variables Y _b . Furthermore, the Gaussian process model is given by a predetermined calculation scheme for an expected value E(X _u ) ∈ ℝ ^{l ×k} and a variance Var(X _u ) for input variables not contained in the original data set for l different operating states X _u ∈ ℝ ^n×l : $$E\left(X_u\right)=m\left(X_u\right)+K\left(X_u,X_b\right){\left(K\left(X_b,X_b\right)+{\sigma }^{2}I\right)}^{-1}\left(Y_b-m\left(X_b\right)\right),$$ $$Var\left(X_u\right)=K\left(X_u,X_u\right)+{\sigma }^2-K\left(X_u,X_b\right){\left(K\left(X_b,X_b\right)+{\sigma }^{2}I\right)}^{-1}K\left(X_b,X_u\right),$$ with a mean function m(X _u ), a predetermined variance σ ² , the identity matrix I, and a covariance function K, which depends on the Euclidean distance r between two points x _1, x ₂ in the following way: $$K\left(X^1,X^2\right)={\left(k\left(X_{1:\text{n:i}}^1,X_{1:\text{n}\text{,j}}^2\right)\right)}_{i=\mathrm{1,}\dots ,m,j=\mathrm{1,}\dots ,m},$$ $$k\left(x_1,x_2\right)={\sigma }_F{}^2\text{exp}\left(-\frac{r{\left(x_1,x_2\right)}^2}{2l^2}\right),$$ with a predetermined distance parameter l and a predetermined signal variance σ _F . Thus, in equations (4) and (5) K(X _u ,X _b ) ∈ ℝ ^l×m , K(X _b ,X _b ) ∈ ℝ ^m×m ,I ∈ ℝ ^m×m and Y _b ∈ ℝ ^m×k .

The mean function m(x) is preferably obtained as a Gaussian process model.

In one embodiment, the nominal model is first calculated as a first Gaussian process model. First input variables X _b1 ⊂ X _b and associated first output variables Y _b1 ⊂ Y _b are selected to determine the first Gaussian process model, in particular an expectation value E _N (X) and a variance Var _N (X) of the nominal model. Furthermore, for the purpose of determining the first Gaussian process model, it is preferably assumed that m(x) = 0 is assumed for its mean function, so that, starting from equations (4) and (5), the equations $$E_N\left(X_u\right)=K\left(X_u,X_{b1}\right){\left(K\left(X_{b1},X_{b1}\right)+{\sigma }^{2}I\right)}^{-1}{Y}_{b1},$$ $$Va{r}_N\left(X_u\right)=K\left(X_u,X_u\right)+{\sigma }^2-K\left(X_u,X_{b1}\right){\left(K\left(X_{b1},X_{b1}\right)+{\sigma }^{2}I\right)}^{-1}K\left(X_{b1},X_u\right),$$ Particularly preferably, the first input variables X _b1 and the first output variables Y _b1 are obtained from a steady-state operation of the power device.

The resulting expectation value E _N (X) of the first Gaussian process model is then used in a next step as the mean function m(x) in a second Gaussian process model, which includes second input variables X _b2 ⊂ X _b and corresponding second output variables Y _b2 ⊂ Y _b , with the second Gaussian process model being used as the detailed model. Based on equations (4) and (5), the equations $$E_D\left(X_u\right)=E_N\left(X_u\right)+K\left(X_u,X_{b2}\right){\left(K\left(X_{b2},X_{b2}\right)+{\sigma }^{2}I\right)}^{-1}\left(Y_{b2}-E_N\left(X_{b2}\right)\right),$$ $$Va{r}_D\left(X_u\right)=K\left(X_u,X_u\right)+{\sigma }^2-K\left(X_u,X_{b2}\right){\left(K\left(X_{b2},X_{b2}\right)+{\sigma }^{2}I\right)}^{-1}K\left(X_{b2},X_u\right).$$

Particularly preferably, the second input variables X _b2 and the second output variables Y _b2 are obtained from a transient operation of the power device.

In one embodiment, the Gaussian process model according to equations (4) to (7), in particular the second Gaussian process model according to equations (6), (7), (10) and (11), is adapted during operation of the power device, especially during transient operation. For this purpose, newly measured data points can be used during operation of the power device. $\left(X_b^{\text{'}},Y_b^{\text{'}}\right)$ supplemented, or data points (X _b ,Y _b ) identified as capable of improvement - in particular (X _b2 ,Y _b2 ) - by newly measured data points $\left(X_b^{\text{'}},Y_b^{\text{'}}\right)$ The newly measured data points $\left(X_b^{\text{'}},Y_b^{\text{'}}\right)$ the majority of reconstruction measured values and the majority of control values, whereby for the newly measured data points $$\left(X_b^{\text{'}},Y_b^{\text{'}}\right)=\left(X_{\left[t_A,t_E\right]},Y_{\left[t_A,t_E\right]}^r\right)$$ applies.

In one embodiment, a model-based predictive control method for the operation of the power device is carried out on the basis of the power device model.

According to a further development of the invention, it is provided that, based on the plurality of control values, a plurality of temporally successive simulation values of the at least one measured variable are determined using the power device model, in particular using the nominal model. The plurality of measured values are manipulated using the plurality of simulation values, the plurality of control values, and the plurality of measured values to obtain the plurality of reconstruction measured values. The simulation values are not influenced by the dynamic behavior of the at least one sensor, since the simulation values are preferably calculated based on the nominal model and thus based on stationary operating data, so that the dynamic behavior of the at least one sensor is advantageously taken into account when manipulating the measured values based on the reconstruction measured values.

In the context of the present technical teaching, a simulation value is preferably understood to be a value of the measured variable determined by means of the power device model. Furthermore, a simulation value is denoted below by the symbol ${\overrightarrow{y}}^s\left(t\right)$ , where t denotes the time of the corresponding control value. Furthermore, regardless of the number of measured variables, the simulation value—preferably represented in vector form—is referred to.

In the predetermined time interval [t _A , t _E ], for the majority of times t _n with t _A ≤ t _n ≤ t _E a simulation value ${\overrightarrow{y}}^s\left(t_n\right)$ This means that the majority of simulation values in the predetermined time interval are stored in a matrix of the form $$Y^s{}_{\left[t_A,t_E\right]}={\left({\overrightarrow{y}}^s\left(t_n\right)\right)}_{n=\mathrm{1,}\dots ,k}$$ shown.

In one embodiment, the majority of reconstruction measurements are calculated using the equation $$Y^r{}_{\left[t_A,t_E\right]}=f\left(X_{\left[t_A,t_E\right]},Y_{\left[t_A,t_E\right]},Y^s{}_{\left[t_A,t_E\right]}\right)$$ calculated, where the function f models the dynamic behavior of at least one sensor.

Preferably, the majority of simulation values are calculated using equation (4) as $$Y^s{}_{\left[t_A,t_E\right]}=E\left(X_{\left[t_A,t_E\right]}\right)=m\left(X_{\left[t_A,t_E\right]}\right)+K\left(X_{\left[t_A,t_E\right]},X_b\right){\left(K\left(X_b,X_b\right)+{\sigma }^{2}I\right)}^{-1}\left(Y_b-m\left(X_b\right)\right)$$ Additionally, a variance of the simulation value is optionally calculated using equation (5) as $$Var\left(X_{\left[t_A,t_E\right]}\right)=K\left(X_{\left[t_A,t_E\right]},X_{\left[t_A,t_E\right]}\right)+{\sigma }^2-K\left(X_{\left[t_A,t_E\right]},X_b\right){\left(K\left(X_b,X_b\right)+{\sigma }^{2}I\right)}^{-1}K\left(X_b,X_{\left[t_A,t_E\right]}\right)$$ calculated.

In one embodiment, the plurality of temporally successive simulation values of the at least one measured variable are determined using the nominal model of the power device model based on the plurality of control values. For this purpose, the plurality of simulation values are preferably determined using equation (8) as $$Y^s{}_{\left[t_A,t_E\right]}=E_N\left(X_{\left[t_A,t_E\right]}\right)=K\left(X_{\left[t_A,t_E\right]},X_{b1}\right){\left(K\left(X_{b1},X_{b1}\right)+{\sigma }^{2}I\right)}^{-1}{Y}_{b1}$$ In addition, a variance of the majority of simulation values is optionally calculated using equation (9) as $$\begin{array}{c}Va{r}_N\left(X_{\left[t_A,t_E\right]}\right)=K\left(X_{\left[t_A,t_E\right]},X_{\left[t_A,t_E\right]}\right)+{\sigma }^2\\ -K\left(X_{\left[t_A,t_E\right]},X_{b1}\right){\left(K\left(X_{b1},X_{b1}\right)+{\sigma }^{2}I\right)}^{-1}K\left(X_{b1},X_{\left[t_A,t_E\right]}\right)\end{array}$$ calculated.

According to a further development of the invention, a sensor dead time and/or a sensor's time response is calculated based on the power device model. The majority of measured values are manipulated based on the sensor dead time and/or the time response. Advantageously, the dynamic behavior of the at least one sensor is easily taken into account using the sensor dead time and/or the time response.

In the context of the present technical teaching, a sensor dead time is preferably understood to mean a time interval between a measurement of a measured variable at the sensor and the recording of the measured value of the measured variable in a control device. Alternatively or additionally, in the context of the present technical teaching, a sensor dead time is also understood to mean a dead time of the measured variable, i.e., in particular, a dead time that occurs between the occurrence or initiation of a change and the effect of the change on the measured variable. For example, to detect nitrogen oxide formation in a combustion chamber of an internal combustion engine, a nitrogen oxide quantity is measured using a sensor in an exhaust path of the internal combustion engine, so that the sensor dead time here corresponds to the travel time of an exhaust gas containing the nitrogen oxides from the combustion chamber to the sensor. The sensor dead time depends on a position of the at least one sensor relative to the control device, with the sensor dead time typically increasing proportionally with a length of a signal conductor that connects the at least one sensor to the control device for data transmission. Therefore, the control value of the at least one control variable and the measured value of the at least one measured variable are shifted in time by the sensor dead time. Furthermore, the sensor dead time is denoted below by the symbol T, where an i-th sensor dead time T _i is assigned to an i-th sensor for measuring the i-th measured variable y _i .

In the context of the present technical teaching, a time response is understood to mean, in particular, a PT1-filtered response. A PT1-filtered response of the sensor is understood to mean that the sensor exhibits a proportional response with a first-order delay when detecting a measured variable. Furthermore, the time response is denoted below by the symbol $\overrightarrow{PT}$ where an i-th time response PT _i is assigned to an i-th sensor for measuring the i-th measured variable y _i .

Particularly preferably, the sensor dead time and/or the time behavior is determined based on the plurality of simulation values and the plurality of measured values, in particular by comparing the plurality of simulation values with the plurality of measured values.

In a preferred embodiment, the majority of reconstruction measurement values are calculated taking into account the sensor dead time using the equation $$Y^r{}_{\left[t_A,t_E\right]}=f\left(X_{\left[t_A,t_E\right]},\overrightarrow{T}\right)$$ determined, in particular $\overrightarrow{T}=g\left(Y_{\left[t_A,t_E\right]},Y^s{}_{\left[t_A,t_E\right]}\right)$ and using the function g the majority of simulation values and the majority of measured values are compared to determine the sensor dead time. Alternatively, the reconstruction measurements are calculated taking into account the time behavior using the equation $$Y^r{}_{\left[t_A,t_E\right]}=f\left(X_{\left[t_A,t_E\right]},\overrightarrow{PT}\right)$$ determined, with preference given to $\overrightarrow{PT}=h\left(Y_{\left[t_A,t_E\right]},Y^s{}_{\left[t_A,t_E\right]}\right)$ and using the function h, the majority of simulation values and the majority of measured values are compared to determine the time response. Alternatively, the reconstruction measured values are calculated taking into account the sensor dead time and the time response using the equation $$Y_{\left[t_{A,}{t}_E\right]}^r=f\left(X_{\left[t_A,t_E\right]},\overrightarrow{T},\overrightarrow{PT}\right)=f\left(X_{\left[t_A,t_E\right]},g\left(Y_{\left[t_A,t_E\right]}^S\right),h\left(Y_{\left[t_A,t_E\right]},Y_{\left[t_A,t_E\right]}^S\right)\right)$$ determined.

According to a further development of the invention, it is provided that a control variable parameter selected from a group consisting of a power device speed, a fuel introduction quantity, a fuel introduction time, a fuel introduction pressure, a power device torque, a charge air pressure, an air mass quantity, an exhaust gas recirculation rate, a combustion air ratio, and a combination of at least two of the said control variable parameters is used as the at least one control variable.

In one embodiment, the manipulated variable consists of the power device speed, the fuel injection quantity, the fuel injection timing, the fuel injection pressure, the charge air pressure, and the air mass quantity. Thus, the manipulated variable vector has six components—in particular, six lines.

According to a further development of the invention, it is provided that a measurement parameter selected from a group consisting of a NO _x amount, an exhaust gas temperature, a particle concentration, a CO amount, a quantity of unburned hydrocarbons, a combustion chamber pressure, a combustion air mass flow, a combustion air pressure, a combustion air temperature, a compressor speed, and a combination of at least two of the said measurement parameters is used as the at least one measurement parameter.

In a preferred embodiment, the measured variable consists of the NO _x amount, the exhaust gas temperature, the particulate concentration, the CO amount, the amount of unburned hydrocarbons, the combustion chamber pressure, the combustion air mass flow, the combustion air pressure, the combustion air temperature, and the compressor speed. Thus, the measured variable vector has ten components—in particular, ten lines.

Preferably, the NO _x amount is measured and/or determined using a NO _x sensor. Alternatively or additionally, the exhaust gas temperature is measured and/or determined using an exhaust gas temperature sensor. Alternatively or additionally, the particle concentration is measured and/or determined using a particle sensor. Alternatively or additionally, the CO amount is measured and/or determined using a CO sensor. Alternatively or additionally, the amount of unburned hydrocarbons is measured and/or determined using a hydrocarbon sensor. Alternatively or additionally, the combustion chamber pressure is measured and/or determined using a combustion chamber pressure sensor. Alternatively or additionally, the combustion air mass flow is measured and/or determined using a combustion air mass sensor. Alternatively or additionally, the combustion air pressure is measured and/or determined using a combustion air pressure sensor. Alternatively or additionally, the combustion air temperature is measured and/or determined using a combustion air temperature sensor. Alternatively or additionally, the compressor speed is measured and/or determined using a speed sensor.

According to a further development of the invention, the method is carried out cyclically. Advantageously, the power device model is continuously adapted, thereby continuously reducing a model error of the power device model.

In one embodiment, the method is performed at a first predetermined frequency. Preferably, the power device model, in particular the detailed model, is adapted at the first predetermined frequency.

In one embodiment, the method is carried out with a first predetermined frequency of at most 1 Hz, wherein the detailed model is adapted with this first predetermined frequency.

According to a further development of the invention, the sensor dead time and/or the time response are determined once or cyclically. Advantageously, the determination of the reconstruction measured values is continuously adjusted, thereby continuously reducing the model error of the power device model.

In one embodiment, the sensor dead time is determined at a second predetermined frequency.

In one embodiment, the sensor dead time and/or the time response is determined using a second predetermined frequency of at most 1/60 Hz. Thus, the sensor dead time and/or the time response advantageously appropriately takes into account a dynamic excitation of the at least one sensor.

According to a further development of the invention, it is provided that the second predetermined frequency is at most as large as the first predetermined frequency.

According to a further development of the invention, it is provided that the predetermined time interval has a duration of 10 seconds to 150 seconds, preferably of 30 seconds to 100 seconds, preferably of 50 seconds to 80 seconds.

According to a further development of the invention, after an adaptation of the power device model, the power device is operated in an evaluation time interval with the plurality of control values. In this case, a plurality of model values of the at least one measured variable are calculated by means of the power device model—in particular by means of the detailed model—based on the plurality of control values. Furthermore, a second plurality of measured values are acquired by means of the sensor in the evaluation time interval, wherein the second plurality of measured values are manipulated by means of the power device model, taking into account the dynamic behavior of the at least one sensor, thereby obtaining a second plurality of reconstruction measured values. Subsequently, the model error of the power device model—in particular of the detailed model—is determined based on the plurality of model values and the second plurality of reconstruction measured values, wherein the power device model is evaluated based on the model error.

In the context of the present technical teaching, a model value is preferably understood to be a value of the measured variable calculated using the power device model. Furthermore, a model value is denoted below by the symbol ${\overrightarrow{y}}^m\left(t\right)$ where t indicates the time point that characterizes the specific period or the specific time of the control value. Furthermore, the model value—especially represented in vector form—is referred to independently of the number of measured variables.

The evaluation time interval is preferably a period from the evaluation start time to an evaluation end time that follows the evaluation start time, where t _BA denotes the evaluation start time and t _BE denotes the evaluation end time. Furthermore, the predetermined time interval and the evaluation time interval are of equal length, so that t _BE - t _BA = t _E - t _A.

In the evaluation time interval [t _BA , t _BE ], at the majority of times t _n with t _BA ≤ t _n ≤ t _BE a model value ${\overrightarrow{y}}^m\left(t_n\right)$ Therefore, the majority of model values in the evaluation time interval are stored in a matrix of the form $$Y_{\left[t_{BA},t_{BE}\right]}^m={\left({\overrightarrow{y}}^m\left(t_n\right)\right)}_{n=\mathrm{1,}\dots ,k}$$ Furthermore, for the majority of control values in the evaluation time interval X _{[t A ,t E ]} = X _{[t BA ,t BE ]} . The second plurality of measured values in the evaluation time interval is denoted by Y _{[t BA ,t BE ]} and the second plurality of reconstruction measurements with $Y_{\left[t_{BA},t_{BE}\right]}^r$ designated.

In one embodiment, the model error is calculated using the equation $$M\left(Y_{\left[t_{BA},t_{BE}\right]}^m,Y_{\left[t_{BA},t_{BE}\right]}^r\right)=\Vert Y_{\left[t_{BA},t_{BE}\right]}^m-Y_{\left[t_{BA},t_{BE}\right]}^r\Vert$$ where M denotes the model error and ||·|| denotes an arbitrary norm. The smaller the model error, the better the performance device model.

The object is also achieved by providing a method for operating the power device. The power device is operated in the predetermined time interval with a transient operating characteristic, wherein a plurality of temporally successive control values of the at least one control variable of the power device are specified in the predetermined time interval. A sensor records a plurality of temporally successive measured values of the at least one measured variable in the predetermined time interval. Based on the plurality of control values, a plurality of temporally successive simulation values of the at least one measured variable are determined using the power device model. The power device model is subsequently adapted based on the plurality of simulation values, the plurality of control values, and the plurality of measured values. The power device is then operated based on the adapted power device model. Due to the dynamic behavior of the at least one sensor, the measured values recorded at a specific time do not reflect the actual values of the at least one measured variable at the location of the at least one sensor at the specific time. Therefore, the manipulated variable and the measured value at the specific point in time cannot be used as data points for adapting the power device model. Typically, the simulation values are not influenced by the dynamic behavior of the at least one sensor, since the simulation values are preferably calculated based on the nominal model and thus based on steady-state operating data, so that the dynamic behavior of the at least one sensor is advantageously taken into account when adapting the power device model. The power device model is thus advantageously adapted during transient operation of the power device. In addition, optimal manipulated variables of the at least one manipulated variable can also be determined for transient operation of the power device based on the adapted power device model.

Particularly preferably, a Gaussian process model with a nominal model according to equations (6) to (9) and a detailed model according to equations (6), (7), (10) and (11) is used as the power device model. The majority of simulation values are calculated using equation (16). Furthermore, the detailed model is adapted based on the majority of simulation values, the majority of control values and the majority of measured values, wherein the newly measured data points are used for the adaptation. $$\left(X_b^{\text{'}},Y_b^{\text{'}}\right)=\left(X_{\left[t_A,t_E\right]},f\left(X_{\left[t_A,t_E\right]},Y_{\left[t_A,t_E\right]},Y_{\left[t_{BA},t_{BE}\right]}^S\right)\right)$$ In addition, the power device is operated based on the detailed model.

According to a further development of the invention, the plurality of measured values is manipulated using the plurality of simulation values, the plurality of control values, and the plurality of measured values to obtain a plurality of reconstruction measured values. The power device model is then adapted based on the plurality of reconstruction measured values and the plurality of control values. Advantageously, the dynamic behavior of the at least one sensor is taken into account in the reconstruction measured values, so that the reconstruction measured values of the specific time and the control values of the specific time can be used as data points for adapting the power device model.

In one embodiment, the majority of reconstruction measurements are calculated using equation (14).

The object is also achieved by providing a control device which is designed to carry out a method according to the invention or a method according to one or more of the The control device is preferably designed as a computing device, particularly preferably as a computer, or as a control unit, preferably as a control unit of an internal combustion engine. In conjunction with the control device, the advantages already described above in connection with the method for operating the power device are realized in particular.

The control device is preferably configured to operate the power device. In one embodiment, the control device is configured to operate an internal combustion engine, an internal combustion engine-generator combination device, i.e., a genset, a fuel cell, an energy storage device, for example, a battery, an electrolyzer, a data center, or a microgrid, or another controllable or regulatable load on an electrical network.

In one embodiment, the control device is configured to operate, i.e. preferably control, the power device with a control value of the at least one control variable.

In one embodiment, the control device is configured to detect, determine or receive the plurality of measured values of the at least one measured variable.

The object is also achieved by providing a power arrangement comprising a power device, at least one sensor, at least one actuator, and a control device according to the invention or a control device according to one or more of the previously described embodiments. In connection with the power arrangement, the advantages already described above in connection with the method and the control device are realized in particular.

The control device is preferably operatively connected to the power device in order to control the power device. Furthermore, the control device is operatively connected to the at least one sensor and the at least one actuator.

Preferably, the at least one sensor is configured to measure the plurality of measured values of the at least one measured variable and transmit them to the control device, wherein the control device is configured to detect the plurality of measured values. Alternatively or additionally, the at least one actuator is configured to be controlled by means of the plurality of control values of the at least one control variable.

In one embodiment, it is provided that the power device is designed as an internal combustion engine, an internal combustion engine-generator combination device, i.e., a genset, a fuel cell, an energy storage device, for example, a battery, an electrolyzer, a data center or a microgrid, or as another controllable or regulatable load on an electrical network.

The object is also achieved by providing an internal combustion engine with at least one combustion chamber, at least one sensor, at least one actuator, and a control device according to the invention or a control device according to one or more of the previously described embodiments. In connection with the internal combustion engine, the advantages already described above in connection with the method and the control device are realized in particular.

• 1 eine schematische Darstellung eines Ausführungsbeispiels einer Leistungsanordnung,
• 2 eine schematische Darstellung eines ersten Ausführungsbeispiels eines Verfahrens zum Betreiben einer Leistungsvorrichtung in Form eines Flussdiagramms,
• 3 eine schematische Darstellung eines Details des Verfahrens gemäß 2,
• 4 eine schematische Darstellung eines zweiten Ausführungsbeispiels des Verfahrens zum Betreiben der Leistungsvorrichtung in Form eines Flussdiagramms, und
• 5 eine schematische Darstellung eines Details des Verfahrens gemäß 4.

The invention is explained in more detail below with reference to the drawings, which show:

• 1 a schematic representation of an embodiment of a power arrangement,
• 2 a schematic representation of a first embodiment of a method for operating a power device in the form of a flow chart,
• 3 a schematic representation of a detail of the process according to 2 ,
• 4 a schematic representation of a second embodiment of the method for operating the power device in the form of a flow chart, and
• 5 a schematic representation of a detail of the process according to 4 .

1 shows a schematic representation of an embodiment of an internal combustion engine 1 with at least one combustion chamber 3 and an embodiment of a control device 5.

The internal combustion engine 1 additionally has a fuel valve 7, an intake valve 9.1, an exhaust valve 9.2, and at least one sensor 11. The fuel valve 7 is arranged in a combustion air path 13—also called a charge air path—along a combustion air flow direction 15—also called a charge air flow direction—fluctually upstream of the intake valve 9.1. The fuel valve 7, the intake valve 9.1, and the exhaust valve 9.2 are designed as actuators.

Optionally, the internal combustion engine 1 also has a turbocharger device 17 with a turbine 19 and a compressor 21. The turbine 19 is arranged in an exhaust gas path 23 along an exhaust gas flow direction 25, fluidically downstream of the exhaust valve 9.2. Furthermore, the compressor 21 is arranged in the combustion air path 13 along the combustion air flow direction 15, fluidically upstream of the fuel valve 7.

Optionally, the internal combustion engine 1 also has an exhaust gas recirculation device 27 with an exhaust gas recirculation path 29 and an exhaust gas recirculation flap 31 designed as an actuator, wherein the exhaust gas recirculation path 29 fluidically connects the exhaust gas path 23 to the combustion air path 13. Furthermore, the exhaust gas recirculation flap 31 is configured to block and/or release the exhaust gas recirculation path 29.

The control device 5 is preferably configured to operate the internal combustion engine at a power device speed, a power device torque, a boost air pressure, an air mass quantity, and/or a combustion air ratio. Alternatively or additionally, the control device 5 is operatively connected to the fuel valve 7 in a manner not explicitly shown and configured to control it in order to adjust a fuel introduction quantity, a fuel introduction time, and/or a fuel introduction pressure. Alternatively or additionally, the control device 5 is operatively connected to the exhaust gas recirculation flap 31 in a manner not explicitly shown and configured to control it in order to adjust an exhaust gas recirculation rate.

Particularly preferably, the at least one sensor 11 is selected from a group consisting of a NO _x sensor 11.1, an exhaust gas temperature sensor 11.2, a particle sensor 11.3, a CO sensor 11.4, a hydrocarbon sensor 11.5, a combustion chamber pressure sensor 11.6, a combustion air mass sensor 11.7, a combustion air pressure sensor 11.8, a combustion air temperature sensor 11.9, a rotational speed sensor 11.10, and a combination of at least two of the aforementioned sensors 11.

The control device 5 is connected to the NO _x sensor 11.1 in a manner not explicitly shown in such a way that a NO _x quantity measured and/or determined by the NO _x sensor 11.1 is transmitted from the NO _x sensor 11.1 to the control device 5. Alternatively or additionally, the control device 5 is connected to the exhaust gas temperature sensor 11.2 in a manner not explicitly shown in such a way that an exhaust gas temperature measured and/or determined by the exhaust gas temperature sensor 11.2 is transmitted from the exhaust gas temperature sensor 11.2 to the control device 5. Alternatively or additionally, the control device 5 is connected to the particle sensor 11.3 in a manner not explicitly shown for data transmission such that a particle concentration measured and/or determined by the particle sensor 11.3 is transmitted from the particle sensor 11.3 to the control device 5. Alternatively or additionally, the control device 5 is connected to the CO sensor 11.4 in a manner not explicitly shown for data transmission such that a CO quantity measured and/or determined by the CO sensor 11.4 is transmitted from the CO sensor 11.4 to the control device 5. Alternatively or additionally, the control device 5 is connected to the hydrocarbon sensor 11.5 in a manner not explicitly shown for data transmission such that a quantity of unburned hydrocarbons measured and/or determined by the hydrocarbon sensor 11.5 is transmitted from the hydrocarbon sensor 11.5 to the control device 5. Alternatively or additionally, the control device 5 is connected to the combustion chamber pressure sensor 11.6 in a manner not explicitly shown in such a way that a combustion chamber pressure measured and/or determined by the combustion chamber pressure sensor 11.6 is transmitted from the combustion chamber pressure sensor 11.6 to the control device 5. Alternatively or additionally, the control device 5 is connected to the combustion air mass sensor 11.7 in a manner not explicitly shown in such a way that a combustion air mass flow measured and/or determined by the combustion air mass sensor 11.7 is transmitted from the combustion air mass sensor 11.7 to the control device 5. Alternatively or additionally, the control device 5 is connected to the combustion air pressure sensor 11.8 in a manner not explicitly shown in such a way that a combustion air pressure measured and/or determined by the combustion air pressure sensor 11.8 is transmitted from the combustion air pressure sensor 11.8 to the control device 5. Alternatively or additionally, the Control device 5 is connected to the combustion air temperature sensor 11.9 in a manner not explicitly shown for data transmission such that a combustion air temperature measured and/or determined by the combustion air temperature sensor 11.9 is transmitted from the combustion air temperature sensor 11.9 to the control device 5. Alternatively or additionally, the control device 5 is connected to the speed sensor 11.10 in a manner not explicitly shown for data transmission such that a compressor speed measured and/or determined by the speed sensor 11.10 is transmitted from the speed sensor 11.10 to the control device 5.

In addition, the control device 5 is configured to carry out a method for operating the power arrangement 1. The method is described in the 2 to 5 presented using flow charts.

2 shows a schematic representation of a first embodiment of a method for operating a power device, for example the internal combustion engine 1, in the form of a flow chart.

Identical and functionally identical elements are provided with the same reference numerals in all figures, so that reference is made to the previous description.

In a step S1 _, a plurality of _temporally successive control values X _{[t A ,t E ]} of at least one manipulated variable of the power device. In one embodiment, the predetermined time interval [t _A , t _E] has a duration of 10 seconds to 150 seconds, preferably 30 seconds to 100 seconds, preferably 50 seconds to 80 seconds.

Preferably, a manipulated variable parameter selected from a group consisting of the power device speed, the fuel introduction quantity, the fuel introduction time, the fuel introduction pressure, the power device torque, the charge air pressure, the air mass quantity, the exhaust gas recirculation rate, the combustion air ratio, and a combination of at least two of the aforementioned manipulated variable parameters is used as the at least one manipulated variable.

In a step S2, the power device is controlled in the predetermined time interval [t _A , t _E ] based on the plurality of control values X _{[t A ,t E ]} in a transient operating sequence.

In a step S3, a plurality of temporally successive measured values Y _{[t A ,t E ]} of at least one measured variable in the predetermined time interval [t _A , t _E] . Preferably, each control value of the plurality of control values X _{[t A ,t E ]} one of the majority of measured values Y _{[t A ,t E ]} assigned.

Preferably, a measurement parameter selected from a group consisting of the NO _x amount, the exhaust gas temperature, the particle concentration, the CO amount, the amount of unburned hydrocarbons, the combustion chamber pressure, the combustion air mass flow, the combustion air pressure, the combustion air temperature, the compressor speed, and a combination of at least two of the said measurement parameters is used as the at least one measurement parameter.

In a step S4, the plurality of measured values Y _{[tA, tE]} are manipulated by means of a power device model GP taking into account a dynamic behavior of the at least one sensor 11, whereby a plurality of temporally successive reconstruction measured values $Y_{\left[t_A,t_E\right]}^r$ Preferably, each control value of the plurality of control values X _{[t A ,t E ]} one reconstruction measurement value of the majority of reconstruction measurements $Y_{\left[t_A,t_E\right]}^r$ assigned.

The power device model GP preferably has a nominal model and a detailed model. Particularly preferably, a Gaussian process model according to equations (4) to (7) is used as the power device model GP, with a first Gaussian process model according to equations (6) to (9) being used as the nominal model and a second Gaussian process model according to equations (6), (7), (10), and (11) being used as the detailed model.

In a step S5, the power device model GP is calculated based on the plurality of reconstruction measurement values $Y_{\left[t_A,t_E\right]}^r$ and the majority of control values X _{[t A ,t E ]} is preferably based on the Majority of reconstruction measurements $Y_{\left[t_A,t_E\right]}^r$ and the majority of control values X _{[t A ,t E ]} the detailed model of the power device model GP—the second Gaussian process model—is adapted. Particularly preferably, the detailed model of the power device model GP is adapted using newly measured data points, with equation (12) applying to the newly measured data points.

In a step S6, the power device is operated based on the adapted power device model GP.

Particularly preferably, steps S1 to S6 are performed cyclically. In this case, the power device model GP, in particular the detailed model, is adapted with a first predetermined frequency of at most 1 Hz.

3 shows a schematic representation of step S4 of the method according to 2 .

In an optional step S7, a plurality of temporally successive simulation values is determined based on the plurality of control values X _[tA,tE] $Y_{\left[t_A,t_E\right]}^S$ the at least one measured variable is determined by means of the power device model GP, in particular by means of the nominal model. Preferably, the plurality of simulation values $Y_{\left[t_A,t_E\right]}^S$ calculated using equation (15). Particularly preferred is the majority of simulation values $Y_{\left[t_A,t_E\right]}^S$ calculated using the nominal model based on equation (17).

In an optional step S8, the power device model GP, preferably the plurality of measured values Y _{[t A ,t E ]} and the majority of simulation values $Y_{\left[t_A,t_E\right]}^S$ a sensor dead time $\overrightarrow{T}$ of the at least one sensor 11 and/or a time behavior $\overrightarrow{PT}$ of at least one sensor 11. Particularly preferably, the sensor dead time $\overrightarrow{T}$ and/or the time behavior $\overrightarrow{PT}$ once or cyclically, preferably with a second predetermined frequency of at most 1/60 Hz.

In an optional step S9, the majority of measured values Y _{[t A ,t E ]} using the majority of simulation values $Y^s{}_{\left[t_A,t_E\right]},$ the majority of control values X _{[t A ,t E ]} and the majority of measured values Y _{[t A ,t E ]} manipulated to obtain the majority of reconstruction measurements $Y^r{}_{\left[t_A,t_E\right]}$ Preferably, the majority of reconstruction measurements $Y^r{}_{\left[t_A,t_E\right]}$ calculated using equation (14). Particularly preferred is the majority of measured values Y _{[t A ,t E ]} based on the sensor dead time $\overrightarrow{T}$ and/or the time behavior $\overrightarrow{PT}$ manipulated, with the majority of reconstruction measurements $Y^r{}_{\left[t_A,t_E\right]}$ is calculated using equation (14) and/or one of equations (17).

4 shows a schematic representation of a second embodiment of the method for operating the power device in the form of a flow chart.

The steps S1 to S3, S6 and S7 are analogous to 2 and 3 formed, whereby step S7 is mandatory.

In a step S10, the power device model GP is calculated based on the plurality of simulation values $Y^s{}_{\left[t_A,t_E\right]},$ the majority of control values X _{[t A ,t E ]} and the majority of measured values Y _{[t A ,t E ]} is particularly preferably adjusted. The detailed model of the power device model GP is adjusted using newly measured data points, whereby equation (20) applies to the newly measured data points.

5 shows a schematic representation of step S10 of the method according to 4 .

The steps S5, S8 and S9 are analogous to 2 and 3 formed, these steps being optional in the method according to 4 are provided.

Overall, it can be seen that the first embodiment of the method, which is 2 and 3 and which has all steps S1 to S10 as mandatory steps, is identical to the second embodiment of the method, which according to 4 and 5 and which has all steps S1 to S10 as mandatory steps.

Claims (13)

Method for operating a power device, wherein - the power device is operated in a predetermined time interval ([t _A , t _E ]) with a transient operating characteristic, wherein - a plurality of temporally successive control values (X _{[t A ,t E ]} ) at least one control variable $\left(\overrightarrow{x}\right)$ of the power device in the predetermined time interval ([t _A , t _E ]), wherein - by means of at least one sensor (11) a plurality of temporally successive measured values (Y _{[t A ,t E ]} ) at least one measured value $\overrightarrow{y}$ in the predetermined time interval ([t _A , t _E ]), wherein - the plurality of measured values (Y _{[t A ,t E ]} ) are manipulated by means of a power device model (GP) taking into account a dynamic behavior of the at least one sensor (11), whereby a plurality of temporally successive reconstruction measured values $\left(Y^r{}_{\left[t_A,t_E\right]}\right)$ ) of at least one measured value $\overrightarrow{y}$ is obtained, wherein - the power device model (GP) is calculated from the plurality of reconstruction measurements $\left(Y^r{}_{\left[t_A,t_E\right]}\right)$ and the majority of control values (X _{[t A ,t E ]} ), whereby - the power device is operated based on the adapted power device model (GP). Procedure according to Claim 1 , where - based on the majority of control values (X _{[t A ,t E ]} ) a plurality of temporally consecutive simulation values $\left(Y^s{}_{\left[t_A,t_E\right]}\right)$ of at least one measured value $\left(\overrightarrow{y}\right)$ is determined by means of the power device model (GP), in particular by means of the nominal model, wherein - the majority of measured values (Y _{[t A ,t E ]} ) using the majority of simulation values $\left(Y^s{}_{\left[t_A,t_E\right]}\right),$ the majority of control values (X _{[t A ,t E ]} ) and the majority of measured values (Y _{[t A ,t E ]} ) to obtain the majority of reconstruction measurements $\left(Y^r{}_{\left[t_A,t_E\right]}\right)$ to obtain. Method according to one of the preceding claims, wherein a sensor dead time is determined based on the power device model (GP) $\left(\overrightarrow{T}\right)$ of the at least one sensor (11) and/or a time behavior $\left(\overrightarrow{PT}\right)$ of the at least one sensor (11), and wherein the plurality of measured values (Y _{[t A ,t E ]} ) based on the sensor dead time $\left(\overrightarrow{T}\right)$ and/or the time behavior $\left(\overrightarrow{PT}\right)$ be manipulated, preferably the sensor dead time (T) and/or the time behavior $\left(\overrightarrow{PT}\right)$ based on the majority of simulation values $\left(Y^s{}_{\left[t_A,t_E\right]}\right)$ and the majority of measured values (Y _{[t A ,t E ]} ), in particular by comparing the majority of simulation values $\left(Y^s{}_{\left[t_A,t_E\right]}\right)$ with the majority of measured values (Y _{[t A ,t E ]} ), is determined. Method according to one of the preceding claims, wherein the at least one manipulated variable $\left(\overrightarrow{x}\right)$ a manipulated variable parameter selected from a group consisting of a power device speed, a fuel injection quantity, a fuel injection timing, a fuel injection pressure, a power device torque, a charge air pressure, an air mass quantity, an exhaust gas recirculation rate, a combustion air ratio, and a combination of at least two of said manipulated variable parameters is used. Method according to one of the preceding claims, wherein the at least one measured variable $\left(\overrightarrow{y}\right)$ a measured variable parameter selected from a group consisting of a NO _x quantity, an exhaust gas temperature, a particle concentration, a CO quantity, a quantity of unburned hydrocarbons, a combustion chamber pressure, a combustion air mass flow, a combustion air pressure, a combustion air temperature, a compressor speed, and a combination of at least two of the said measured variable parameters is used. Method according to one of the preceding claims, wherein the method is carried out cyclically, in particular with a first predetermined frequency, wherein preferably the power device model (GP), in particular the detailed model, is adapted with the first predetermined frequency. Method according to one of the Claims 3 until 7 , where the sensor dead time $\left(\overrightarrow{T}\right)$ and/or the time behavior $\left(\overrightarrow{PT}\right)$ once or cyclically, in particular with a second predetermined frequency. Procedure according to Claim 6 and 7 , wherein the second predetermined frequency is at most as large as the first predetermined frequency. Method according to one of the preceding claims, wherein the predetermined time interval ([t _A , t _E ]) has a duration of 10 seconds to 150 seconds, in particular of 30 seconds to 100 seconds, in particular of 50 seconds to 80 seconds. Method for operating a power device, wherein - the power device is operated in a predetermined time interval ([t _A , t _E ]) with a transient operating characteristic, wherein - a plurality of temporally successive control values (X _{[t A ,t E ]} ) at least one control variable $\left(\overrightarrow{x}\right)$ of the power device in the predetermined time interval ([t _A , t _E ]), wherein - by means of at least one sensor (11) a plurality of temporally successive measured values (Y _{[t A ,t E ]} ) at least one measured value $\left(\overrightarrow{y}\right)$ in the predetermined time interval ([t _A , t _E ]), whereby - based on the plurality of control values (X _{[t A ,t E ]} ) a plurality of temporally consecutive simulation values $\left(Y_{\left[t_A,t_E\right]}^S\right)$ of at least one measured value $\left(\overrightarrow{y}\right)$ is determined by means of a power device model (GP), wherein - the power device model (GP) is determined based on the plurality of simulation values $\left(Y_{\left[t_A,t_E\right]}^S\right),$ the majority of control values (X _{[t A ,t E ]} ) and the majority of measured values (Y _{[t A ,t E ]} ), whereby - the power device is operated based on the adapted power device model (GP). Control device (5) configured to carry out a method according to one of the preceding claims. Power arrangement with a power device, a sensor (11), an actuator and a control device (5) according to Claim 11 . Internal combustion engine (1) with a combustion chamber (3), a sensor (11), an actuator and a control device (5) according to Claim 11 .