JP6483571B2 - Performance estimation apparatus and performance estimation method - Google Patents

Description

  The present invention relates to a performance estimation device and a performance estimation method.

  In a gas turbine engine or the like, there are internal state quantities and performance parameters that cannot be measured by sensors. There is a technique using a Kalman filter as a technique for estimating values of internal state quantities and performance parameters that cannot be measured by such sensors. For example, Patent Document 1 describes a performance estimation system for a gas turbine engine that calculates an unobservable variable estimated value from a sensor measurement value using a Kalman filter including a physical model of the engine.

Japanese Patent No. 5046104

  In the performance estimation system described in Patent Document 1, the Kalman filter gain is constant. Since this Kalman filter gain is obtained at a certain operating point of the gas turbine engine, at the operating point deviating from the operating point from which the Kalman filter gain is obtained, the estimation accuracy of the state quantity that cannot be measured by the sensor is lowered and cannot be estimated. In some cases.

  The present invention provides a performance estimation device and a performance estimation method capable of improving the estimation accuracy of a state quantity that cannot be measured by a sensor.

  A performance estimation device according to one aspect of the present invention is based on a sensor input value based on a control input value for controlling a target device that is an estimation target device and a sensor measurement value measured by a sensor provided in the target device. This is a device that outputs a state estimated value that is an estimated value of the state quantity of the target device that cannot be used. This performance estimation device uses a nonlinear simulation model of the target device, based on a control input value, an estimated value before adjusting an estimation error of a state estimated value, and an estimated value of a sensor measurement value Based on an estimation unit that calculates a certain sensor estimation value, an error calculation unit that calculates an error between the sensor estimation value calculated by the estimation unit and the sensor measurement value, an error calculated by the error calculation unit, and a robust filter gain And an adjustment unit that generates an adjustment value for causing the estimation unit to adjust the prior estimation value. The estimation unit adjusts the prior estimation value based on the adjustment value generated by the adjustment unit, calculates the state estimation value, and outputs the calculated state estimation value.

  In this performance estimation device, the pre-estimated value and the sensor estimated value are calculated using the nonlinear simulation model of the target device, and the adjustment value generated based on the error of the sensor estimated value and the sensor measured value and the robust filter gain Based on the above, the prior estimated value is adjusted and the state estimated value is calculated. As described above, by using the nonlinear simulation model of the target device, the estimation accuracy is improved as compared with the case of using the linear simulation model. In addition, in the robust filter, compared to the Kalman filter, the error between the simulation model and the target device to be estimated, and the change in the state of the target device due to disturbance factors such as a change in the operating point of the target device, The estimated state value can be calculated, and the stability of performance estimation can be improved. As a result, it is possible to improve the estimation accuracy of the performance of the target device that cannot be measured by the sensor.

  The robust filter gain may be a fixed value calculated so that the estimation error falls within a predetermined range. In this case, since the robust filter gain is a fixed value, it is not necessary to update the robust filter gain. Thereby, it is possible to reduce the calculation load.

  A performance estimation method according to another aspect of the present invention is a method of outputting a state estimated value that is an estimated value of a state quantity that cannot be measured by a sensor provided in a target device that is an estimation target device. This performance estimation method uses a step of receiving a control input value for controlling the target device, a step of receiving a sensor measurement value measured by a sensor provided in the target device, and a nonlinear simulation model of the target device. Calculating a pre-estimated value that is an estimated value before adjusting the estimation error of the state estimated value based on the control input value, and a sensor estimated value that is an estimated value of the sensor measurement value; Calculating an error from the sensor measurement value, generating an adjustment value for adjusting the pre-estimation value based on the error and the robust filter gain, and adjusting the pre-estimation value based on the adjustment value. Calculating the state estimated value and outputting the state estimated value.

  In this performance estimation method, the pre-estimated value and the sensor estimated value are calculated using the nonlinear simulation model of the target device, and the adjustment value generated based on the error of the sensor estimated value and the sensor measured value and the robust filter gain. Based on the above, the prior estimated value is adjusted and the state estimated value is calculated. As described above, by using the nonlinear simulation model of the target device, the estimation accuracy is improved as compared with the case of using the linear simulation model. In addition, in the robust filter, compared to the Kalman filter, the error between the simulation model and the target device to be estimated, and the change in the state of the target device due to disturbance factors such as a change in the operating point of the target device, The estimated state value can be calculated, and the stability of performance estimation can be improved. As a result, it is possible to improve the estimation accuracy of the performance of the target device that cannot be measured by the sensor.

  According to the present invention, it is possible to improve the estimation accuracy of a state quantity that cannot be measured by a sensor.

It is a figure which shows schematic structure of the performance estimation system containing the performance estimation apparatus which concerns on one Embodiment. It is a figure which shows typically each element of the gas turbine engine of FIG. It is a hardware block diagram of the performance estimation apparatus of FIG. It is a flowchart which shows a series of processes of the performance estimation method which the performance estimation apparatus of FIG. 1 performs. It is a figure which shows schematic structure of the performance estimation system containing the performance estimation apparatus of a comparative example.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the description of the drawings, the same elements are denoted by the same reference numerals, and redundant descriptions are omitted.

  FIG. 1 is a diagram illustrating a schematic configuration of a performance estimation system including a performance estimation device according to an embodiment. A performance estimation system 1 shown in FIG. 1 is a system that estimates the state of a target device based on sensor measurement values measured by a plurality of sensors provided in the target device that is an estimation target device. Hereinafter, a gas turbine engine will be described as an example of the target device. In addition, when a value at a certain time k of a variable that changes with time is expressed, there may be a case where (k) is added to the code representing the variable, but the description is not limited to a specific time, and is established at an arbitrary time. . The unit time of the performance estimation system 1 is “1”, and the larger the value in parentheses, the longer the time has elapsed. Also, in the figures and mathematical formulas, Δ may be added to the sign representing each variable, which means the amount of change from the value of the variable at the operating point at which the later-described robust filter gain K is calculated. In the following description, when both the value and the amount of change are valid, the description will be made using the value.

  The performance estimation system 1 includes a gas turbine engine 10 and a performance estimation device 20. The gas turbine engine 10 is used as an engine of a vehicle, a ship, an aircraft, etc., for example.

  FIG. 2 is a diagram schematically showing each element of the gas turbine engine 10. As shown in FIG. 2, the gas turbine engine 10 includes a fan (FAN) 11, a low pressure compressor (LPC) 12, a high pressure compressor (HPC) 13, a combustor (COMB) 14, and a high pressure turbine ( HPT) 15, low-pressure turbine (LPT) 16, rotor 17, and rotor 18.

  The fan 11 sucks air from the outside of the gas turbine engine 10 and supplies a part of the sucked air to the low-pressure compressor 12. The low pressure compressor 12 compresses the air supplied from the fan 11 to generate compressed air, and supplies the compressed air to the high pressure compressor 13. The high pressure compressor 13 further compresses the compressed air supplied from the low pressure compressor 12 to further increase the pressure to generate high pressure air, and supplies the high pressure air to the combustor 14. The combustor 14 mixes fuel with the high-pressure air supplied from the high-pressure compressor 13 and burns it. The combustor 14 supplies a high-temperature and high-pressure combustion gas obtained by combustion to the high-pressure turbine 15.

  The high-pressure turbine 15 is rotated by the high-temperature and high-pressure combustion gas supplied from the combustor 14 and rotates the rotor 18. The low pressure turbine 16 is rotated by the combustion gas that has passed through the high pressure turbine 15 and rotates the rotor 17. The rotor 17 connects the fan 11, the low-pressure compressor 12, and the low-pressure turbine 16 so as to be integrally rotatable. The rotor 18 connects the high-pressure compressor 13 and the high-pressure turbine 15 so as to be integrally rotatable.

  Returning to FIG. 1, the gas turbine engine 10 is provided with a plurality of sensors (not shown). The gas turbine engine 10 is controlled by an engine controller (not shown). The gas turbine engine 10 inputs a control input value u (k) output from the engine controller at a certain time k, and a sensor measurement value y (k) detected by a sensor provided in the gas turbine engine 10. Is output. The control input value u (k) is a value for controlling the gas turbine engine 10 and is output based on engine operating conditions. When the gas turbine engine 10 is provided in an aircraft, the engine operating conditions are altitude, speed, and temperature date. The control input value u (k) includes the fuel flow rate, the opening degree of the extraction valve, and the like. The sensor measurement value y (k) is r-dimensional and includes the number of rotations of the rotors 17 and 18 of the gas turbine engine 10 and the temperature and pressure at each part.

The performance estimation device 20 estimates the state quantity x (k) of the gas turbine engine 10 by a robust filter using a nonlinear model of the gas turbine engine 10 and a fixed robust filter gain K. The robust filter is, for example, an H∞ filter. The non-linear model is a non-linear simulation model determined in advance for each gas turbine engine 10, and is expressed by the following formulas (1) to (3) using the non-linear function f and the non-linear function g. The nonlinear function f is predetermined for each target device (gas turbine engine 10), and includes an internal state quantity x _e (k−1), a performance parameter value q (k−1), and a control input value u (k). It is a function. The nonlinear function g is predetermined for each target device, and is a function of the internal state quantity x _e (k) and the performance parameter value q (k). The internal state quantity x _e is an amount of the state inside the gas turbine engine 10 including a state quantity that cannot be measured by the sensor. The internal state quantity x _e is a _ne- dimensional vector and includes the rotational speed of each element, the internal energy of each element, the air flow rate of each element, the enthalpy of each element, and the like. The performance parameter value q is an engine performance parameter value that cannot be measured by the sensor. The performance parameter value q is a p-dimensional vector, and includes a fan flow rate, fan efficiency change, compressor flow rate, compressor efficiency change, and the like.

Specifically, the performance estimation device 20 outputs the estimated state value x _h (k) of the gas turbine engine 10 based on the control input value u (k) and the sensor measurement value y (k). The state estimated value x _h (k) is an estimated value of the state quantity x (k) of the gas turbine engine 10. The state quantity x (k) is a state quantity including a state quantity that cannot be measured by the sensor among the state quantities of the gas turbine engine 10, and as will be described later, the internal state quantity x _e (k) and the performance parameter. It is an n (n = n _e + p) dimensional vector combined with the value q (k). The state estimated value x _h (k) is an n (n = n _e + p) -dimensional vector obtained by combining the estimated value of the internal state quantity x _e (k) and the estimated value of the performance parameter value q (k).

  FIG. 3 is a hardware configuration diagram of the performance estimation apparatus 20. As shown in FIG. 3, the performance estimation device 20 physically includes one or a plurality of processors 201, a storage device 202 such as a RAM (Random Access Memory) and a ROM (Read Only Memory) that are main storage devices, It may be configured as a computer including hardware such as an auxiliary storage device 203 such as a hard disk device, an input device 204 such as a keyboard, an output device 205 such as a display, and a communication device 206 that is a data transmission / reception device. Each function shown in FIG. 1 of the performance estimation device 20 is performed under the control of one or more processors 201 by causing hardware such as the processor 201 and the storage device 202 to read one or more predetermined computer programs. This is realized by operating each hardware and reading and writing data in the storage device 202 and the auxiliary storage device 203.

  Returning to FIG. 1, the details of the performance estimation device 20 will be described. The performance estimation device 20 includes an error calculation unit 21, an adjustment unit 22, and an estimation unit 30.

The error calculation unit 21 calculates an error e (k) between the sensor measurement value y (k) output from the gas turbine engine 10 and the sensor estimation value y _h (k) calculated by the estimation unit 30. In this example, the error calculation unit 21 subtracts the sensor estimated value y _h (k) from the sensor measurement value y (k), and outputs the subtraction result to the adjustment unit 22 as an error e (k). The sensor estimated value y _h (k) is an estimated value of the sensor measured value y (k).

Based on the error e (k) calculated by the error calculation unit 21 and the robust filter gain K, the adjustment unit 22 causes the estimation unit 30 to pre-estimate the value x _h ⁻ ( An adjustment value for adjusting k) is generated. The prior estimated value x _h ⁻ (k) is an estimated value before adjusting the estimation error of the state estimated value x _h (k). Specifically, the adjustment unit 22 multiplies the error e (k) output from the error calculation unit 21 and the robust filter gain K, and outputs the multiplication result Ke (k) to the estimation unit 30 as an adjustment value. . The robust filter gain K is a fixed value gain calculated so that the estimation error falls within a predetermined range. The robust filter gain K is calculated off-line and stored in advance in the adjustment unit 22.

Here, a method of calculating the robust filter gain K will be described. First, a nonlinear model of the gas turbine engine 10 is prepared. This nonlinear model is determined in advance for each target device, and the nonlinear model of the gas turbine engine 10 is expressed by the following equations (1) to (3). More specifically, the internal state quantity x _e (k) is expressed as the sum of the nonlinear function f and the system noise v _xe (k) as shown in the equation (1). System noise _v xe (k) is a vector of _{n e} dimensions. The performance parameter value q (k) is expressed as the sum of the performance parameter value q (k−1) and the performance parameter change amount v _q (k) as shown in the equation (2). The performance parameter change amount v _q (k) is a p-dimensional vector. The sensor measurement value y (k) is expressed as the sum of the nonlinear function g and the observation noise w (k), as shown in Equation (3). The observation noise w is an r-dimensional vector.

Subsequently, the nonlinear model of the gas turbine engine 10 is linearized. This linearization is performed with reference to a certain operating point. As the operating point, for example, the maximum output time in a stationary state on the ground is used. The nonlinear model linearized at the operating point includes a change amount Δx (k) from the state quantity x at the operating point, a change amount Δy (k) from the sensor measurement value y at the operating point, and a control input value u at the operating point. Is expressed by, for example, Expression (4) and Expression (5) using the change amount Δu (k), noise v (k), and observation noise w (k). The matrices A, B, and C are determined in advance by the combination of the target device and the operating point.

As shown in Expression (6), the state quantity x is an n (n = n _e + p) -dimensional vector obtained by combining the internal state quantity x _e and the performance parameter value q. The noise v is an n (n = n _e + p) -dimensional vector obtained by combining the system noise v _xe and the performance parameter change amount v _q . The noise v is a finite noise that satisfies the equation (8). The observation noise w is a finite noise that satisfies Equation (9).

Expressions (1) and (2) are transformed using Expressions (6) and (7) to obtain Expression (1) ′. Moreover, Formula (3) is deform | transformed using Formula (6), and Formula (3) 'is obtained.

In the robust filter, the estimation error between the state quantity x (k) of the gas turbine engine 10 and the state estimation value x _h (k) estimated by the estimation unit 30 needs to be within a certain range. Therefore, equation (10) needs to be established using the estimation error index γ. The index γ is set to a value with which a desired estimation accuracy can be obtained and a solution of Expression (10) can be obtained.

Note that x _b (0) is an average value of the initial distribution of the state quantity x. The matrix Q is a weight matrix for the noise v and satisfies the equation (11). The matrix R is a weight matrix for the observation noise w and satisfies the equation (12).

Here, using the evaluation function J shown in Expression (13), the solution of Expression (10) is examined.

Since the matrix Q and the matrix R take positive values, the problem of obtaining the change amount Δx _h of the state estimated value that satisfies the equation (10) under Q> 0 and R> 0 is the change amount of the state estimated value. Δx _h is fixed, which is equivalent to the fact that the evaluation function J <0 holds for the amount of change Δx (0) of the initial value of the arbitrary state quantity x, the noise v, and the observation noise w. Represented as:

Further, the equation (10) obtains the change amount Δx _h of the estimated state value that minimizes the evaluation function J by fixing the change amount Δx (0) of the initial value of the state amount x, the noise v, and the observation noise w. And is expressed as equation (15).

In order for the evaluation function J to have an extreme value in order to satisfy the equation (15), it is necessary that dJ = 0. Therefore, using the covariance P of the estimation error and the unit matrix I, the following riccati Equation (16) is derived.

Specifically, the following equations (17) to (19) are used to obtain the estimation error covariance P based on the equation (16).

By solving equations (17) to (19), a robust filter gain K is obtained as shown in equation (20).

The estimation unit 30 calculates the prior estimated value x _h ⁻ (k) and the sensor estimated value y _h (k) based on the control input value u (k) using the nonlinear model of the gas turbine engine 10. Based on the adjustment value generated by the adjustment unit 22, the estimation unit 30 adjusts the prior estimation value x _h ⁻ (k) to calculate the state estimation value x _h (k), and calculates the calculated state estimation value x _h. (K) is output. The estimating unit 30 outputs the calculated sensor estimated value y _h (k) to the error calculating unit 21 and outputs the calculated state estimated value x _h (k) to the outside. The estimating unit 30 includes a time updating unit 31, an estimated value updating unit 32, an estimated value updating unit 33, and a delay unit 34.

The time updating unit 31 calculates the prior estimated value x _h ⁻ (k) based on the control input value u (k) and the state estimated value x _h (k−1) output from the delay unit 34. . Specifically, the time updating unit 31 calculates the prior estimated value x _h ⁻ (k) by using the nonlinear function f used in Expression (1) and Expression (21). The time updating unit 31 outputs the calculated prior estimated value x _h ⁻ (k) to the estimated value updating unit 32 and the estimated value updating unit 33.

The estimated value updating unit 32 calculates the sensor estimated value y _h (k) based on the prior estimated value x _h ⁻ (k) output from the time updating unit 31. Specifically, the estimated value updating unit 32 calculates the sensor estimated value y _h (k) by using the nonlinear function g used in Expression (3), using Expression (22). The estimated value updating unit 32 outputs the calculated sensor estimated value y _h (k) to the error calculating unit 21.

The estimated value updating unit 33 calculates the state estimated value x _h (k) based on the prior estimated value x _h ⁻ (k) output from the time updating unit 31 and the adjusted value output from the adjusting unit 22. To do. Specifically, the estimated value update unit 33 is an adder, and as shown in Expression (23), the prior estimated value x _h ⁻ (k) and the adjustment value are added, and the addition result is the state estimated value. Calculate as x _h (k). The estimated value update unit 33 outputs the calculated state estimated value x _h (k) to the delay unit 34 and the outside.

The delay unit 34 delays the state estimated value x _h (k) output from the estimated value updating unit 33 by unit time, and updates the time estimated as the state estimated value x _h (k−1) with respect to the estimation at the next time. To the unit 31.

  Next, a series of processes of the performance estimation method performed by the performance estimation device 20 will be described with reference to FIG. FIG. 4 is a flowchart showing a series of processes of the performance estimation method performed by the performance estimation apparatus 20. The process shown in FIG. 4 is repeatedly performed every unit time. Here, processing at a certain time k will be described.

  First, the engine controller outputs a control input value u (k) to the gas turbine engine 10 and the performance estimation device 20 based on the engine operating conditions. Then, the time update unit 31 of the performance estimation device 20 receives the control input value u (k) (step S01). Further, the gas turbine engine 10 receives the control input value u (k) output from the engine controller, and outputs a sensor measurement value y (k) measured by a sensor provided in the gas turbine engine 10. . Then, the error calculation unit 21 of the performance estimation device 20 receives the sensor measurement value y (k) (step S02).

Subsequently, the time updating unit 31 calculates a prior estimated value x _h ⁻ (k) based on the control input value u (k) using a nonlinear model of the gas turbine engine 10 (step S03). Specifically, the time update unit 31 uses the control input value u (k) and the state estimated value x _h (k−1) output from the delay unit 34 to perform the pre-estimation according to the equation (21). The value x _h ⁻ (k) is calculated. Then, the time update unit 31 outputs the calculated prior estimated value x _h ⁻ (k) to the estimated value update unit 32 and the estimated value update unit 33.

Subsequently, the estimated value update unit 32 calculates a sensor estimated value using a nonlinear model of the gas turbine engine 10 (step S04). Specifically, the estimated value updating unit 32 calculates the sensor estimated value y _h (k) using the prior estimated value x _h ⁻ (k) output from the time updating unit 31 according to Expression (22). Then, the estimated value updating unit 32 outputs the calculated sensor estimated value y _h (k) to the error calculating unit 21.

Subsequently, the error calculation unit 21 calculates an error e (k) between the sensor measurement value y (k) and the sensor estimated value y _h (k) (step S05). Specifically, the error calculation unit 21 subtracts the sensor estimation value y _h (k) from the sensor measurement value y (k), and outputs the subtraction result to the adjustment unit 22 as an error e (k).

Subsequently, the adjustment unit 22 generates an adjustment value for adjusting the pre-estimated value x _h ⁻ (k) based on the error e (k) calculated by the error calculation unit 21 and the robust filter gain K. (Step S06). Specifically, the adjustment unit 22 multiplies the error e (k) and the robust filter gain K, and outputs the multiplication result Ke (k) to the estimation unit 30 as an adjustment value.

Subsequently, the estimated value updating unit 33 adjusts the prior estimated value x _h ⁻ (k) based on the adjustment value to calculate the state estimated value x _h (k) (step S07). Specifically, as shown in Expression (23), the estimated value update unit 33 adds the prior estimated value x _h ⁻ (k) and the adjustment value, and uses the addition result as the state estimated value x _h (k ). Then, the estimated value updating unit 33 outputs the calculated state estimated value x _h (k) to the outside and the delay unit 34 (step S08). As described above, a series of processes of the performance estimation method at time k ends.

As described above, the performance estimation device 20 performs the state quantity x (k) (internal state quantity x) of the actual machine (gas turbine engine 10) by the robust filter using the nonlinear model of the gas turbine engine 10 and the fixed robust filter gain K. _e (k) and performance parameter value q (k)) are estimated. The performance estimation apparatus 20 then determines the error e (k) so that the error e (k) between the sensor measurement value y (k) and the sensor estimation value y _h (k) output from the estimation unit 30 is small. Is multiplied by a robust filter gain K and adjustment of the prior estimated value x _h ⁻ (k) is repeated to estimate the state quantity x (k) of the gas turbine engine 10, that is, the state estimated value x _h (k). Is calculated.

  Next, the effects of the performance estimation device 20 will be described while comparing with the performance estimation system 100 including the performance estimation device 120 of the comparative example. FIG. 5 is a diagram illustrating a schematic configuration of a performance estimation system 100 including a performance estimation apparatus 120 of a comparative example. As shown in FIG. 5, the performance estimation system 100 is different from the performance estimation system 1 in that a performance estimation device 120 is provided instead of the performance estimation device 20.

The performance estimation device 120 is mainly different from the performance estimation device 20 in that the robust filter gain K (k) is not fixed and a linear model is used instead of the nonlinear model of the gas turbine engine 10. That is, the performance estimation device 120 is a device that estimates the state estimated value x _h (k) of the gas turbine engine 10 using the linear model of the gas turbine engine 10 and the robust filter gain K (k), and an error calculation unit. 121, an adjustment unit 122, and an estimation unit 130.

  The linear model is expressed by the following expressions (24) and (25) using the matrices A, B, and C. The matrices A, B, and C are predetermined for each operating point for each target device (gas turbine engine 10). The matrices A, B, and C are the same as the matrices A, B, and C in the equations (4) and (5).

  A series of processes of the performance estimation method performed by the performance estimation device 120 will be described. First, the engine controller outputs a control input value u (k) to the gas turbine engine 10 and the performance estimation device 120 based on the engine operating conditions. Then, the time update unit 131 of the performance estimation device 120 receives the control input value u (k). Further, the gas turbine engine 10 receives the control input value u (k) output from the engine controller, and outputs a sensor measurement value y (k) measured by a sensor provided in the gas turbine engine 10. . Then, the error calculation unit 121 of the performance estimation device 120 receives the sensor measurement value y (k).

Subsequently, the time updating unit 131 calculates a prior estimated value x _h ⁻ (k) based on the control input value u (k) using the linear model of the gas turbine engine 10. Specifically, the time update unit 131 includes a calculation unit 131a, a calculation unit 131b, and an adder 131c. First, the operation unit 131a multiplies the matrix A and the state estimation value x _h (k−1) output from the delay unit 134, and outputs the multiplication result to the adder 131c. In addition, the calculation unit 131b performs multiplication of the matrix B and the control input value u (k), and outputs the multiplication result to the adder 131c. Then, the adder 131c adds the multiplication result of the calculation unit 131a and the multiplication result of the calculation unit 131b, and calculates the addition result as a prior estimated value x _h ⁻ (k). That is, the time update unit 131 calculates the prior estimated value x _h ⁻ (k) by performing the calculation shown in Expression (24). Then, the time updating unit 131 outputs the calculated prior estimated value x _h ⁻ (k) to the estimated value updating unit 132 and the estimated value updating unit 133.

Subsequently, the estimated value update unit 132 calculates a sensor estimated value using a linear model of the gas turbine engine 10. Specifically, the estimated value update unit 132 performs multiplication of the matrix C and the prior estimated value x _h ⁻ (k) output from the time update unit 131 as shown in Expression (25), and the multiplication result Is calculated as a sensor estimated value y _h (k). Then, the estimated value updating unit 132 outputs the calculated sensor estimated value y _h (k) to the error calculating unit 121.

Subsequently, similarly to the error calculation unit 21, the error calculation unit 121 calculates an error e (k) between the sensor measurement value y (k) and the sensor estimated value y _h (k). Specifically, the error calculation unit 121 subtracts the sensor estimation value y _h (k) from the sensor measurement value y (k), and outputs the subtraction result to the adjustment unit 122 as an error e (k).

Subsequently, the adjusting unit 122 adjusts the prior estimated value x _h ⁻ (k) based on the error e (k) calculated by the error calculating unit 121 and the robust filter gain K (k). Is generated. Specifically, as shown in Expression (26), the adjustment unit 122 first has a unit matrix I, a matrix C, a weighting matrix R for the observation noise w, an estimation error index γ, and an estimation error covariance P. The matrix Ψ (k) is calculated using (k−1). Then, as shown in Expression (27), the adjustment unit 122 uses the matrix A, the covariance P (k−1) of the estimation error, the calculated matrix Ψ (k), and the weight matrix Q for the noise v. The covariance P (k) of the estimation error is calculated. Further, the adjustment unit 122 calculates the robust filter gain K (k) using the calculated estimation error covariance P (k), the matrix C, and the weighting matrix R, as shown in Expression (28). Then, the adjustment unit 122 multiplies the error e (k) output from the error calculation unit 121 and the robust filter gain K (k), and uses the multiplication result K (k) e (k) as an adjustment value for the estimation unit. To 130.

Subsequently, similarly to the estimated value updating unit 33, the estimated value updating unit 133 is based on the prior estimated value x _h ⁻ (k) output from the time updating unit 131 and the adjustment value output from the adjusting unit 122. Then, the estimated state value x _h (k) is calculated. Specifically, the estimated value update unit 133 is an adder, adds the pre-estimated value x _h ⁻ (k) and the adjustment value as shown in Expression (29), and uses the addition result as the state estimated value. Calculate as x _h (k). Then, the estimated value update unit 133 outputs the calculated state estimated value x _h (k) to the delay unit 134 and the outside. As described above, a series of processes of the performance estimation method at time k performed by the performance estimation device 120 is completed.

  Thus, since the performance estimation apparatus 120 uses the linear model of the gas turbine engine 10, the estimation accuracy of the nonlinear gas turbine engine 10 decreases. Further, in the performance estimation device 120, it is necessary to prepare a linear model corresponding to the operating point, and it is necessary to update the robust filter gain K (k) every unit time, so that the calculation load is high.

On the other hand, the performance estimation device 20 calculates the prior estimated value x _h ⁻ (k) and the sensor estimated value y _h (k) using the nonlinear model of the gas turbine engine 10, and the sensor estimated value y _h (k). Based on the adjustment value generated based on the error e (k) of the sensor measurement value y (k) and the robust filter gain K, the pre-estimated value x _h ⁻ (k) is adjusted to obtain the state estimated value x _h ( k) is calculated. Thus, in the performance estimation device 20, the estimation accuracy is improved by using the nonlinear model of the gas turbine engine 10 as compared with the case where the linear model is used. Further, the robust filter gain K is a fixed value calculated so that the estimation error falls within a predetermined range. For this reason, it is not necessary to update the robust filter gain K. Thereby, it is possible to reduce the calculation load.

Further, in the robust filter, as compared with the Kalman filter, errors in the simulation model and the gas turbine engine 10 and changes in the state of the gas turbine engine 10 due to disturbance factors such as changes in the operating point of the gas turbine engine 10 are caused. Regardless, the estimated state value x _h (k) can be calculated, and the stability of performance estimation can be improved. That is, the stability of performance estimation at operating points other than the filter design point is improved.

  As mentioned above, although embodiment of this invention was described, this invention is not limited to the said embodiment. For example, in the above embodiment, a gas turbine engine is used as an example of the target device, but the present invention is not limited thereto. The target device is a device that performs a non-linear operation with respect to the control input value u (k). The target input is the control input value u (k), and the sensor measurement value y (k) and the state quantity x (k) Any device that can be expressed by a non-linear simulation model as an output may be used.

  In the above embodiment, the adjustment unit 22 uses a fixed-value robust filter gain K. However, the adjustment unit 22 may calculate a robust filter gain K (k) for each unit time. In this case, the adjusting unit 22 linearizes the nonlinear model at the operating point at each time point as shown in the equations (4) and (5), and based on the equation (16) (specifically, the equation The covariance P of the estimation error is calculated using (17) to (19). Then, the adjustment unit 22 calculates a robust filter gain K (k) using Expression (20).

DESCRIPTION OF SYMBOLS 10 Gas turbine engine 20 Performance estimation apparatus 21 Error calculation part 22 Adjustment part 30 Estimation part e Error K Robust filter gain u Control input value x State quantity x _h State estimated value x _h ^- Prior estimated value y Sensor measured value y _h Sensor estimation value

Claims (3)

Based on the control input value for controlling the target device that is the device to be estimated and the sensor measurement value measured by the sensor provided in the target device, the state quantity of the target device that cannot be measured by the sensor A performance estimation device that outputs a state estimation value that is an estimation value,
Based on the control input value using a non-linear simulation model of the target device, a pre-estimated value that is an estimated value before adjusting an estimation error of the state estimated value, and a sensor that is an estimated value of the sensor measurement value An estimation unit for calculating an estimated value;
An error calculating unit that calculates an error between the sensor estimated value calculated by the estimating unit and the sensor measured value;
An adjustment unit that generates an adjustment value for causing the estimation unit to adjust the prior estimation value based on the error calculated by the error calculation unit and a robust filter gain;
With
The estimation unit adjusts the prior estimation value based on the adjustment value generated by the adjustment unit, calculates the state estimation value, and outputs the calculated state estimation value.   The performance estimation apparatus according to claim 1, wherein the robust filter gain is a fixed value calculated so that an estimation error falls within a predetermined range. A performance estimation method for outputting a state estimated value that is an estimated value of a state quantity that cannot be measured by a sensor provided in a target device that is an estimation target device,
Receiving a control input value for controlling the target device;
Receiving a sensor measurement value measured by a sensor provided in the target device;
Based on the control input value using a non-linear simulation model of the target device, a pre-estimated value that is an estimated value before adjusting an estimation error of the state estimated value, and a sensor that is an estimated value of the sensor measurement value Calculating an estimate;
Calculating an error between the sensor estimated value and the sensor measurement value;
Generating an adjustment value for adjusting the prior estimated value based on the error and a robust filter gain;
Calculating the state estimated value by adjusting the prior estimated value based on the adjusted value;
Outputting the estimated state value;
A performance estimation method including:

Images