JPS63143608A - Identification device - Google Patents

Abstract

PURPOSE:To perform the identification of a system parameter with high accuracy by calculating two identification values based on the deformation of a model type corresponding to an object system and obtaining an identification value by means of factor conversion and averaging processing. CONSTITUTION:The titled device is constituted of a 1st identification computing element 3 for calculating the identification value of a factor of a model type corresponding to an object system, a 2nd identification computing element 4 for calculating the identification value of a factor of a model type obtained by replacing a term on the left side of the model type by an optional term on the right side and an average processor 5 for obtaining the average value between the output of the 1st computing element 3 and that of the 2nd computing element 4. When no factor is included in the left sides of two model types, e.g. a 1st model type corresponding to the object system and a 2nd model type obtained by replacing a term on the left side of the 1st model type by a term on the right side, the factors of the terms on both right sides are in inverse relation each other. When the average value between the inverse of the identification value of the factor in the 2nd model type and the identification value of the factor in the 1st model type is adopted, an identification error can be reduced.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to an identification device, and particularly to an identification device that performs coefficient identification calculations of a model equation.

(Prior Art) In various industrial systems such as electric power systems, it is becoming common to use model equations that define the characteristics of the control systems of these systems to perform adaptive control and fault diagnosis of the systems. In such cases, identification of the coefficients of the above model equation, so-called system parameters and parameters, is the basis for understanding the characteristics of these systems and performing adaptive control, fault diagnosis, etc. An example of such an identification device is shown in, for example, Japanese Patent Laid-Open No. 159912/1983.

(Problems to be Solved by the Invention) However, when identification calculations are performed using a conventional identification device, observation noise causes identification errors when observing the state values of the system to be identified. This observation noise increases the identification error especially when attempting to identify using a small identification excitation signal because the S/N ratio is poor. In cases such as online identification of an actual system, identification must be performed using a small identification excitation signal in order to suppress negative effects on the system. Therefore, in these cases, the influence of noise is even more problematic, and high precision Identification was difficult.

The present invention is a method (7) made to solve the above problems.
−? It is an object of the present invention to provide an identification device that can identify system parameters with high accuracy.

[Structure of the invention]

(Means for Solving the Problems) The configuration for achieving the above object will be explained with reference to FIG. a second identification calculator 4 that calculates the identification value of a coefficient of a model equation in which a term on the left side and an arbitrary term on the right side are exchanged in the model equation of The average processor 5 calculates the average value of the output of the second identification calculator 3 and the output of the second identification calculator 4.

(Operation) According to the present invention, a first model equation corresponding to the target system and a second model equation in which the terms on the left side and one term on the right side of this model equation are exchanged are used. If no coefficient is attached to the left-hand side, the coefficients of the terms on the right-hand side have a reciprocal relationship with each other. Therefore, if the average value of the reciprocal of the identified value of the coefficient in the second model equation and the identified value of the coefficient in the first model equation is taken, the identification error becomes smaller.

(Basic Idea of the Invention) As the idea of the invention, the relationship between the identified excitation signal and the identified value of the coefficient will be explained using equations.

First, when the model equation is given by the following equation (1), the identification coefficient value using the least squares method is determined by the equation (2).

y(kl=A −z(kl+v(kl
... (1) ygR', zgR", AgRlx"
A: Identification target coefficient matrix y, z: State value V: Regular white noise k: Sampling step Here, consider the case where the identification excitation signal y falls and approaches O.
Each term of the matrix A based on equation (2) also approaches O. Therefore, in general, unless other factors are considered, when the excitation signal for identification in the term on the left side decreases, the identified value of the coefficient also decreases. However, even if only one term in 2 on the right side has an excitation signal, all terms in the matrix will not become O unless the excitation signal on the left side y does not decrease. Therefore, if one term of y on the left side and 2 on the right side is swapped, the identification accuracy of the two model formulas before and after the swap will be different, but as mentioned above, the two model formulas for the replaced variable 2. Identification accuracy can be improved by using the reciprocal relationship of the coefficients. In short, for example, when the identification excitation signal decreases, if the identified value of the relevant coefficient in the first model equation decreases, the reciprocal of the identified value of the relevant coefficient in the second model equation increases, so these are averaged. This is because the processing cancels out identification errors.

(Example) An example will be described below with reference to the drawings. FIG. 1 is a block diagram of an embodiment of an identification device according to the present invention.

The control system in Figure 1 consists of a control system 1 and a controlled system 2.
It consists of the state value z(k) of the control system,
An identification device (e) is provided which inputs y(k) as an identification excitation signal to obtain an arithmetic device. The identification device 3 includes a first identification calculator 3 that calculates identification values of coefficients of a first model equation, and a second model in which a term on the left side of the first model equation and an arbitrary term on the right side are exchanged. a second identification calculator 4 that calculates identification values of coefficients of the equation, the two identification calculators 3;
An average processor 5 that performs averaging processing between the identified values outputted from 4.
Therefore, the calculation output 6'f! :,It has gained.

An embodiment will be described below, taking as an example a case where the model equation of the target system is expressed by the relationship between the driving torque of the rotating body, the load, and the inertial force of the rotating body.

At this time, the first and second model equations are as follows (3) # (
4) Equation becomes.

2,'5~・7・However, θ: Rotation angle t: Load t: Drive torque IL, & 2: Coefficient The term on the left side of the first model equation (3) and the second term on the right side are swapped. This is the second model equation (4). (3).

As can be seen from equation (4), the coefficients of the second term on the right side of both equations have a reciprocal relationship with each other.

In FIG. 2, the state value transition 1.1 in the transient state of the system shows that the system reaches an equilibrium state after passing through the transient state.

FIG. 3 is a diagram showing identification coefficients in the case of FIG. 2. The identified values of the coefficients 12 and 1/a2 decrease due to the attenuation of the respective state values. At this time, the reciprocal of 1/a2 is a2/. Therefore, what is finally output as the identification value is IL2 after the averaging process.

For the averaging process, either an arithmetic average or a geometric average is used depending on the target system.

Remaining coefficient 1. For, the identified value of the coefficient a by the formula (3) and the identified value of the coefficient IL, 7112 by the formula (4) are multiplied by the previously obtained coefficient 12' to obtain a,', and the coefficient a, The final identified value a, is obtained by averaging the identified values of a,'.

As a result, the identified values a1 and 1□ of the coefficients are smaller than when model equations (3) and (4) are identified alone, respectively.
Identification errors can be reduced.

FIG. 4 is a flowchart for explaining the operation of the averaging processor.

First, in order to obtain the identified value a2, the reciprocal of the coefficient 1/&2 is taken in the stellar f41 to give 82', and in step 42, the coefficient &2/ and the coefficient a2 are averaged. On the other hand, to obtain the identified values a, k, in step 43 the coefficient jL1/a2 is multiplied by a2'e to obtain a, /, and in step 44, the coefficients a, ' and the coefficient a are averaged.

FIG. 5 is a block diagram of another embodiment of the present invention.

In this example, when a transient state occurs in the identified excitation input,
The identification calculation is stopped to prevent the identification error from increasing. 8 is a transient state determiner that detects the occurrence and attenuation of the transient state of the identified excitation input, and uses the identification calculation start/stop signal S' (H). If there is an extreme decrease, the identification calculation is stopped.Also, if it is used to start and stop automatically, the identification value can be calculated every time a transient state occurs and can be applied to online monitoring of the target system.

〔Effect of the invention〕

As explained above, according to the present invention, two identified values are calculated based on the transformation of the model formula according to the target system, and the identified values are obtained by coefficient conversion and averaging processing. It is possible to correct the identification error that occurs when the identified excitation signal is attenuated due to identification, etc., and
In this case, a parameter identification device that can be used online can be provided without considering the addition of special hardware or the addition of a new identification excitation signal.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the identification device according to the present invention, FIG. 2 is a transition diagram of state values in a transient state of the system, FIG. 3 is a diagram showing identification coefficients in the case of FIG. 2, and FIG. FIG. 4 is a flow chart for explaining the operation of the average processor, and FIG. 5 is a block diagram of another embodiment of the present invention. 1... Control system 2... Controlled system 3...
First identification calculator 4...Second identification calculator 5...
Average processor 6...Calculation output ヱ...Identification device

Claims (2)

[Claims] (1) In an identification device that calculates identified values of coefficients of a model formula according to a target system, a first identification calculator that calculates identified values of coefficients of the model formula; a second identification calculator that calculates the identification value of the coefficient of the model equation by replacing an arbitrary term on the right side; and an average value of the output of the first identification calculator and the output of the second identification calculator. An identification device characterized by comprising an average processor for calculating. (2) The identification device according to claim 1, wherein each of the first and second identification calculators is operated in response to a start/stop signal from a transient state determiner.