KR102686946B1 - Apparatus and method for estimating parameters of SPMSM driving system using NLMS adaptive filter - Google Patents

Abstract

The present invention discloses an apparatus and method for estimating SPMSM driving system parameters using an NLMS adaptive filter.
The present invention estimates the electrical parameters of the SPMSM driving system online using the NLMS algorithm, and compared to the conventional method of electrical parameter estimation using adaptive filters such as RLS (Recursive Least Square) and EKF (Extended Kalman Filter), It has the lowest computational complexity and lowest memory requirements, resulting in the lowest CPU load rate when applied to actual systems.
In addition, other conventional adaptive filter algorithm-based parameter estimation methods operate as one of the entire motor drive system applications, but due to their inherent computational complexity and memory requirements, they can be operated by incorporating them into the current control loop that must operate at the fastest frequency. However, the electrical parameter estimation method using the NLMS adaptive filter of the present invention has a small impact on the operation of the entire motor drive system even when operated at the same frequency as the current control loop, so it has various usability in online parameter estimation and electrical It has the effect of being able to estimate parameters in real time under normal motor control conditions.

Description

Apparatus and method for estimating parameters of SPMSM driving system using NLMS adaptive filter {Apparatus and method for estimating parameters of SPMSM driving system using NLMS adaptive filter}

The present invention relates to a parameter estimation device and method, and more specifically, to an estimation device and method for estimating parameters of a SPMSM driving system using an adaptive filter based on the NLMS (Normalized Least Mean Square) algorithm.

SPMSM (Surface-mounted Permanent Magnet Synchronous Motor) is widely used in various application fields such as electric vehicles and air conditioners along with drives for variable speed control. This is because SPMSM has the characteristics of high control precision, power efficiency, simple structure, small load, and high-speed dynamic response. In a high-performance SPMSM control system, information about accurate system parameters becomes an important factor in determining the control precision of the entire system. .

However, in the case of parameters of the entire system, especially electrical parameters, the nominal value includes errors caused by factors such as system temperature, magnetization saturation, and system aging, making it difficult to construct a high-performance control system. In order to do this, it must be possible to estimate these system parameters at the moment of need and apply the estimated values to the system. Typical electrical parameters that require such estimation in SPMSM control systems include winding resistance, inductance, and rotor flux linkage.

Methods for estimating these electrical parameters are divided into online estimation methods and offline estimation methods. The offline parameter estimation method is a method performed under specific conditions and times when closed-loop control of speed or torque is not in progress, and the stator resistance value is determined by temperature. It can be measured using a measurement module, or a method such as measuring the rotor magnetic flux linkage using the stator inductance value measured by applying active current directly offline based on a magnetization circuit model is used. Sometimes it happens.

Even though offline parameter estimation techniques can accurately estimate motor parameters, these methods have the disadvantage of making it difficult to reflect real-time changing parameter values in control and incurring additional costs due to additional interface configuration. do.

On the other hand, the method of estimating electrical parameters online generally uses the SPMSM dynamic state equation on the dq-axis coordinate system rotating at synchronous speed, and separate drive stopping even when controlling speed or torque, which is the purpose of most motor drive systems in industrial systems. Parameters can be estimated in real time. However, since the online parameter estimation method uses state equations, to estimate more than two parameters at a time, the rank deficiency of the system matrix of the SPMSM state equation must be overcome.

In this case, because the dq-axis stator current equation has a low order of the system matrix compared to the three parameters to be estimated, the prior art considers one or two parameters as constant parameters represented by nominal values, so the number of parameters to be estimated is A method of reducing the number was used. Because of this, there was a limitation in that only parameters excluding values considered as nominal values could be estimated.

The problem to be solved by the present invention is to provide an estimation device and method that can accurately estimate the electrical parameters of a surface-mounted permanent magnet synchronous motor (SPMSM) driving system at the required time without increasing costs due to additional interface configuration.

The SPMSM drive system parameter estimation device using an NLMS adaptive filter according to a preferred embodiment of the present invention to solve the above-mentioned problems receives the rotational speed, current, and voltage of the motor from the motor drive system and calculates the estimated value of the stator inductance of the motor. a first estimation unit that generates and outputs an estimated value of the stator inductance to the motor driving system; And receiving the estimated value of the stator inductance from the first estimation unit, receiving the rotation speed, current, and voltage of the motor from the motor driving system, generating an estimated stator resistance value and a rotor flux linkage estimated value of the motor to drive the motor. It includes a second estimator that outputs output to the system, and the first estimator and the second estimator are implemented as adaptive filters based on a Normalized Least Mean Square (NLMS) algorithm.

In addition, when the first estimation unit organizes the continuous-time stator side voltage-current state equation of a surface-mounted permanent magnet synchronous motor (SPMSM) on a dq coordinate system rotating at synchronous speed with the stator inductance as a variable, the stator inductance is calculated as NLMS It is set as the estimation target system of the algorithm, the coefficients combined with the stator inductance are set as input values to the estimation target system of the NLMS algorithm, and the remaining terms not combined with the stator inductance are set as the output values of the estimation target system to obtain the stator inductance. An estimate of can be generated.

In addition, the first estimation unit can estimate the stator inductance in a state where the motor driving system sets the d-axis current to 0 and controls the SPMSM (Surface-Mounted Permanent Magnet Synchronous Motor) with the q-axis current as a command value. there is.

In addition, when the second estimation unit organizes the continuous-time stator side voltage-current state equation of a surface-mounted permanent magnet synchronous motor (SPMSM) on a dq coordinate system rotating at synchronous speed with the stator resistance and rotor flux linkage as variables, , the stator resistance and rotor flux linkage are set as the estimation target system of the NLMS algorithm, the coefficients coupled to the stator resistance and rotor flux linkage are set as input values to be input as the estimation target system of the NLMS algorithm, and the stator resistance and rotor flux linkage are set as input values for the estimation target system of the NLMS algorithm. The remaining terms that are not combined with the electromagnetic flux linkage can be set as the output values of the estimation target system to generate estimated values of the stator resistance and rotor flux linkage.

Additionally, the second estimator may estimate stator resistance and rotor flux linkage while injecting a d-axis current less than 0 for a certain period of time while the motor driving system drives the motor in a normal state.

Meanwhile, the SPMSM driving system parameter estimation method using an NLMS adaptive filter according to a preferred embodiment of the present invention to solve the above-described problem is a method of estimating the electrical parameters of a motor in a parameter estimation device, including (a) the parameter estimation A device receiving the rotational speed, current, and voltage of a motor from a motor driving system, generating an estimated stator inductance of the motor, and outputting the estimated stator inductance to the motor driving system; and (b) using the estimated value of the stator inductance estimated in step (a) and the rotational speed, current, and voltage of the motor supplied from the motor driving system, to calculate the estimated stator resistance and rotor flux linkage value of the motor. It includes the step of generating and outputting to the motor driving system, wherein steps (a) and (b) are performed using an adaptive filter based on the Normalized Least Mean Square (NLMS) algorithm.

In addition, in step (a), when organizing the continuous-time stator side voltage-current state equation of a surface-mounted permanent magnet synchronous motor (SPMSM) on a dq coordinate system rotating at synchronous speed with the stator inductance as a variable, the stator inductance is It is set as the estimation target system of the NLMS algorithm, the coefficients combined with the stator inductance are set as input values to the estimation target system of the NLMS algorithm, and the remaining terms not combined with the stator inductance are set as the output values of the estimation target system to estimate the stator. An estimate of the inductance can be generated.

In addition, step (a) is to estimate the stator inductance in a state where the motor driving system sets the d-axis current to 0 and controls the SPMSM (Surface-Mounted Permanent Magnet Synchronous Motor) with the q-axis current as a command value. You can.

In addition, step (b) is to organize the continuous-time stator side voltage-current state equation of a surface-mounted permanent magnet synchronous motor (SPMSM) on a dq coordinate system rotating at synchronous speed using stator resistance and rotor flux linkage as variables. When, the stator resistance and rotor flux linkage are set as the estimation target system of the NLMS algorithm, the coefficients coupled to the stator resistance and rotor flux linkage are set as input values input to the estimation target system of the NLMS algorithm, and the stator resistance and The remaining terms that are not combined with the rotor flux linkage can be set as the output values of the estimation target system to generate estimated values of the stator resistance and rotor flux linkage.

Additionally, in step (b), the stator resistance and rotor flux linkage can be estimated while injecting a d-axis current smaller than 0 for a certain period of time while the motor driving system drives the motor in a normal state.

The present invention estimates the electrical parameters of the SPMSM driving system online using the NLMS algorithm, and compared to the conventional method of electrical parameter estimation using adaptive filters such as RLS (Recursive Least Square) and EKF (Extended Kalman Filter), It has the lowest computational complexity and lowest memory requirements, resulting in the lowest CPU load rate when applied to actual systems.

The above effect of the present invention is more evident when implementing a motor drive system used in actual industrial sites. The software within the motor drive device for actual industrial sites basically includes a control program that controls speed, torque, and position, and fault handling and A diagnostic program that performs system monitoring is installed as standard, but a program that estimates parameters is installed and operates additionally.

The faster the operating frequency of each program in the motor drive software, the better the responsiveness. To achieve this, a high-performance DSP must be installed in the device, which leads to an unnecessary increase in cost, resulting in a low-cost, high-efficiency motor. It is unsuitable for implementing a driving system.

Therefore, in general, the operating frequency of the current control loop among each program is set at the sharpest level and the remaining programs are set significantly lower than that to secure the operation reliability of the control program, which must be most fundamentally guaranteed.

If the parameter estimation program that needs to be installed additionally operates at the same operating frequency as that of the current control loop and has minimal influence on the basic control program and diagnostic program, high-performance parameter estimation responsiveness can be secured, making its use feasible. It becomes easier.

However, in the case of the parameter estimation method based on the conventional adaptive filter algorithm using the RLS algorithm or Kalman filter algorithm, due to its inherent computational complexity and memory requirements, the operating frequency must be set to a significantly lower operating frequency compared to the operating frequency of the control program to ensure smooth operation of the entire motor drive software. does not affect

On the other hand, in the case of the NLMS adaptive filter-based parameter estimation technique according to the present invention, the computational complexity and memory requirements are significantly lower than those of other adaptive filter algorithm-based parameter estimation methods, so even if the program is driven at the same operating frequency as the control program operating frequency, the entire motor can be driven. Because it causes a low load increase in the software, it is possible to set a higher operating frequency, so compared to other adaptive filter-based parameter estimation techniques, high responsiveness can be expected when applied to an actual system, so its usability is diverse, and electrical parameters can be adjusted to the motor There is an effect that can be estimated in real time in a typical control situation.

Figure 1 is a diagram showing a configuration in which a SPMSM drive system parameter estimation device using an NLMS adaptive filter, a motor, and a motor drive system are connected according to a preferred embodiment of the present invention.
Figure 2 is a diagram showing an example of a general motor drive system according to the prior art.
Figure 3 is a block diagram showing the configuration of a parameter estimation device according to a preferred embodiment of the present invention.
Figure 4 is a diagram illustrating an adaptive filter operation method based on the Normalized Least Mean Square (NLMS) algorithm.
Figures 5 and 6 are graphs comparing performance when parameters are estimated using the NLMS adaptive filter of the present invention and when parameters are estimated using the RLS adaptive filter of the prior art.

The electrical parameter estimation method using the NLMS adaptive filter according to a preferred embodiment of the present invention estimates three parameters by designing two adaptive filters based on the voltage-current state equation on the dq coordinate system rotating at synchronous speed, and the first The NLMS adaptive filter estimates the stator inductance, and the second NLMS adaptive filter estimates the stator resistance and rotor flux linkage.

In addition, of the two NLMS adaptive filters implemented according to the preferred embodiment of the present invention, the first adaptive filter is a general SPMSM MTPA (Maximum Torque Per Ampere) vector control controlled by setting the d-axis current to zero and the q-axis current as the command value. Operation is possible in any situation, and the second adaptive filter operates using data obtained in a situation where the d-axis current is applied as a negative command value for a certain period, but the SPMSM vector control state can be maintained.

In a preferred embodiment of the present invention, the control state of instantaneously applying negative d-axis current does not affect the speed and torque control of the SPMSM drive system including the load, so the electrical parameter estimation method using two adaptive filters can operate online.

Hereinafter, preferred embodiments of the present invention will be described with reference to the attached drawings.

FIG. 1 is a diagram illustrating a configuration in which a SPMSM drive system parameter estimation device 300 using an NLMS adaptive filter, a motor 200, and a motor drive system 100 are connected according to a preferred embodiment of the present invention.

Referring to Figure 1, the SPMSM driving system parameter estimation device 300 (hereinafter abbreviated as "parameter estimation device") using an NLMS adaptive filter according to a preferred embodiment of the present invention is a motor that drives the motor in conjunction with a motor. It is interoperable with the driving system 100.

The motor driving system 100 generates a control command for driving the motor and outputs it to the motor driving unit 120, and generates a voltage to the motor according to the control command input from the control command generating unit 110. It is configured to include a motor drive unit 120 that drives the motor by supplying excess current. In addition, the motor driver 120 measures the physical quantity of the motor and outputs it to the control command generator 110, and the control command generator 110 generates a control command again using the physical quantity input from the motor driver 120. and output to the motor driving unit 120.

The motor drive system 100 shown in FIG. 1 converts the stator side voltage and current into a d-q coordinate system for speed control, and then controls the active and reactive current using signals classified into d-axis component and q-axis component. It is based on a general linear control technique that allows The motor drive system 100 shown in FIG. 1 may selectively adopt specifications suitable for the present invention from among conventional technologies.

For reference, an example of a general motor drive system 100 according to the prior art is shown in FIG. 2. Those skilled in the art will recognize that the motor drive system 100 applied to the present invention is not limited to the example shown in FIG. 2. In addition, since the functions of the control command generator 110 and the motor drive unit 120 shown in FIG. 1 are not significantly different from those of the conventional motor drive system 100, detailed descriptions will be omitted.

Meanwhile, the parameter estimation device 300 of the present invention provides d-axis current (i _d ), d-axis voltage (U _d ), q-axis current (i _q ), and q-axis voltage (U _q ) and The motor rotor angular velocity (ω _e ) is input, and this is used to estimate the stator resistance ( [Ω]), an estimate of the stator inductance ( [H]), an estimate of the rotor flux linkage ( [Wb]) is generated and output to the motor driving system 100 (control command generation unit 110).

First, before describing the parameter estimation device 300 and method according to a preferred embodiment of the present invention, the parameters and variables used in the present invention are defined as follows.

i _d , i _q , U _d , U _q : d-axis current [A], q-axis current [A], d-axis command voltage [V], q-axis command voltage on the stator side on the dq-axis reference coordinate system rotating at synchronous speed. [V]

,

,

: 동기속도로 회전하는 d-q축 기준 좌표계상 고정자 전류 전압 상태 방정식 상의 고정자 저항 [Ω], 고정자 인덕턴스 [H], 회전자 쇄교 자속 [Wb]

,

,

: Stator resistance [Ω], stator inductance [H], rotor magnetic flux linkage [Wb] on the stator current voltage state equation on the dq axis reference coordinate system rotating at synchronous speed

, , : Estimated value of stator resistance [Ω], estimated value of stator inductance [H], estimated value of rotor flux linkage [Wb] on the stator current voltage state equation on the dq axis reference coordinate system rotating at synchronous speed.

ω, ω _e : Mechanical angular velocity of the rotor [rad/s], electrical angular velocity [rad/s]

μ: Step size of NLMS algorithm

,

: NLMS 알고리즘의 적응 이득(Adaptive gain), 정규화 인자(Regularization factor)

,

: Adaptive gain, regularization factor of NLMS algorithm

Meanwhile, assuming that the SPMSM to be controlled has no influence on magnetic saturation due to cross-coupling, structural asymmetry, iron loss, eddy current loss, harmonic components generated by the winding structure, anisotropy of the rotor, and coercive force of the magnet, etc. The continuous-time stator side voltage-current state equation of SPMSM in the dq coordinate system rotating at synchronous speed is as shown in Equation 1.

The parameter estimation device 300 according to a preferred embodiment of the present invention is implemented as an adaptive filter according to the NLMS algorithm, based on Equation 1 above.

Figure 3 is a block diagram showing the configuration of a parameter estimation device 300 according to a preferred embodiment of the present invention.

[Ω]) 및 회전자 쇄교 자속의 추정값(

[Wb])을 생성하는 제 2 추정부(320)를 포함한다.Referring to FIG. 3, the parameter estimation device 300 according to a preferred embodiment of the present invention provides a stator inductance estimate ( ) and a first estimation unit 310 that generates an estimated value of stator resistance (

[Ω]) and an estimate of the rotor flux linkage (

and a second estimation unit 320 that generates [Wb]).

Additionally, both the first estimator 310 and the second estimator 320 are implemented as adaptive filters based on the Normalized Least Mean Square (NLMS) algorithm.

Figure 4 is a diagram illustrating an adaptive filter operation method based on the Normalized Least Mean Square (NLMS) algorithm.

이고, 추정하고자 하는 대상 시스템(Unknown system)이 θ(k)이며, 추정 대상 시스템의 출력신호가 y(k)라고 할 때,

는 추정 대상 시스템 θ(k)를 추정하기 위한 적응 필터 계수가 되고, 는 적응 필터의 출력이 되며, e(k)는 추정 대상 시스템의 출력(y(k))과 적응 필터 출력() 간의 순시 오차가 된다. 이 때, 순시 오차(e(k))가 최소가 되게 하는 적응 필터 계수()가 추정 대상 시스템의 추정값이 된다. 이러한 적응 필터를 이용한 추정 방식은 공지의 사항이므로, 구체적인 설명은 생략한다.Referring to Figure 4, the input signal is

, and the target system to be estimated (unknown system) is θ(k), and the output signal of the estimated target system is y(k),

is the adaptive filter coefficient for estimating the estimation target system θ(k), is the output of the adaptive filter, and e(k) is the output of the estimation target system (y(k)) and the adaptive filter output ( ) is the instantaneous error between At this time, the adaptive filter coefficient ( ) becomes the estimated value of the estimation target system. Since the estimation method using this adaptive filter is well-known, detailed description will be omitted.

Referring again to FIG. 3, the first estimation unit 310 is implemented as an NLMS adaptive filter, and in a general speed control or torque control state of the motor, such as a Maximum Torque Per Ampere (MTPA) control state, the motor driver 120 Receive i _q , U _d , ω _e as input from , and stator inductance ( ) is estimated and output to the second estimation unit 320 and the motor driving system 100.

The first estimation unit 310 calculates the stator inductance ( ) is estimated. The d-axis voltage-current equation in Equation 1 in the steady state where the d-axis current is controlled by the command value is expressed as Equation 2 in order to express it in a form applicable to the NLMS adaptive filter as shown in FIG. 4. do.

은 수학식 2에 의한 출력값으로서 수학식 1의 첫 번째 식의 경우

로 표현되고,

은 입력 신호로서

로 표현되며, θ₁ 은 알수 없는 고정자 인덕턴스 값을 나타내어

가 된다.In Equation 2 above,

is the output value by Equation 2, and in the case of the first equation of Equation 1

It is expressed as,

is the input signal

It is expressed as , where θ ₁ represents the unknown stator inductance value.

It becomes.

At this time, as shown in FIG. 4, the output and error of the NLMS adaptive filter are expressed as Equation 3.

는 NLMS 적응 필터의 출력신호가 되고,

는 θ₁을 추정하기 위한 NLMS 적응 필터의 계수가 되며,

은 NLMS 적응 필터 출력과

간의 순시 오차를 나타냄은 상술한 바와 같다.In Equation 3 above,

becomes the output signal of the NLMS adaptive filter,

is the coefficient of the NLMS adaptive filter for estimating θ ₁ ,

is the NLMS adaptive filter output and

The instantaneous error between intervals is as described above.

)를 아래의 수학식 4에 따라서 추정한다. 즉, 아래의 수학식 4에 따라서

을 구함으로써 고정자 인덕턴스 추정값()을 생성하여 출력한다.The first estimation unit 310 calculates the stator inductance to be estimated according to Equations 2 and 3 (

) is estimated according to Equation 4 below. That is, according to Equation 4 below:

By finding the stator inductance estimate ( ) is generated and output.

은 스텝 사이즈를,

은 적응 이득을,

은 작은 값의 정규화 인자를 각각 나타낸다. 수학식 4의 두 번째 식의 우변에 나타나 있듯이,

은

의 연산 수행 시 분모항이 0이 되는 것을 방지하기 위한 파라미터이며,

는 이산 시간 샘플링 순간의 인덱스를 의미한다.Since the relational expression shown in Equation 4 is a general formula for the NLMS adaptive filter according to the NLMS algorithm, detailed description will be omitted. In Equation 4 above,

is the step size,

is an adaptation benefit,

represents the normalization factor of small values, respectively. As shown on the right side of the second equation in Equation 4,

silver

This is a parameter to prevent the denominator from becoming 0 when performing the calculation.

refers to the index of a discrete time sampling moment.

로 표현되고,

로 표현되며, 가 된다. 이를 수학식 4에 대입하여 에 대해서 풀면, 제 1 추정부(310)는 고정자 인덕턴스 추정값()을 구할 수 있고, 제 1 추정부(310)는 고정자 인덕턴스 추정값()을 제 2 추정부(320) 및 모터 구동 시스템(100)으로 출력한다.The first estimation unit 310 calculates the stator inductance (e.g., MTPA control state) when the motor performs a general control operation. ) is estimated, so at this time, i _d =0. Therefore, in Equation 1 and Equation 4 above,

It is expressed as,

It is expressed as It becomes. By substituting this into equation 4, Solving for , the first estimation unit 310 calculates the stator inductance estimate ( ) can be obtained, and the first estimation unit 310 calculates the stator inductance estimate ( ) is output to the second estimation unit 320 and the motor driving system 100.

Meanwhile, the second estimation unit 320 is also implemented as an NLMS adaptive filter, and receives information from the motor driving unit 120. is input, and the stator inductance estimated value ( ) is input, and the stator resistance is estimated according to Equation 1 above to obtain an estimated stator resistance value ( ), and estimate the rotor flux linkage to obtain the rotor flux linkage estimate ( ) is created.

Assuming that the d-axis and q-axis currents are controlled to be respective command currents, and that the system is in a steady state, the stator inductance estimated by the first estimation unit 310 When using , when the motor drive system 100 is in a normal control state, Equation 1 is expressed as the following Equation 5.

는 입력 행렬을, θ₂는 알 수 없는 파라미터 벡터를 각각 나타내며, 수학식 6에 정의된 바와 같다.At this time, y ₂ is the expected output signal vector,

represents the input matrix and θ ₂ represents the unknown parameter vector, respectively, as defined in Equation 6.

For reference, the second estimation unit 320 performs estimation while injecting a d-axis current id smaller than 0 for a specific time period while the motor driving system 100 drives the motor in a normal state. Therefore, since the rate of change of the d-axis current id and the q-axis current iq is 0, Equation 1 is expressed as Equation 6.

은 SPMSM의 MTPA 제어시 정방 행렬을 구성할 수 없어, 후술하는 수학식 8의 연산에 이용할 수 없다. 따라서, 제 2 추정부(320)가 고정자 저항 추정값()을 생성하고, 회전자 쇄교 자속 추정값()을 생성하기 위해서는, 추가적인 정상상태 운전이 요구된다. 이는 구체적으로 d축 전류 제어 값의 변경을 의미한다. 이때, d 축 전류는 자기 포화 현상을 발생시키지 않도록 음의 상수 값으로 제어되어야 한다. d축 전류 인가시 그 주기는 실제 시스템 구동과 관련된 요구사항에 의해 유동적으로 선택할 수 있다. 다만, 주기 내 획득한 샘플 개수가 필터의 수렴을 보장할 수 있도록 하는 최소 주기 이상으로 d축 전류를 인가하여야 한다. 이러한 최소 주기 이상으로 d축 전류 인가 상태를 유지할 경우, NLMS 적응 필터의 파라미터 식별 수행을 위해 필요한 샘플의 획득과 더불어 NLMS 적응 필터 동작의 기본적인 안정성을 보장할 수 있다. However, the input matrix

cannot form a square matrix during MTPA control of SPMSM and cannot be used in the calculation of Equation 8, which will be described later. Therefore, the second estimation unit 320 calculates the stator resistance estimate value ( ), and the rotor flux linkage estimate ( ), additional steady-state operation is required. This specifically means changing the d-axis current control value. At this time, the d-axis current must be controlled to a negative constant value to avoid magnetic saturation phenomenon. When d-axis current is applied, the period can be flexibly selected according to requirements related to actual system operation. However, the d-axis current must be applied for more than the minimum period that ensures the convergence of the filter by the number of samples acquired within the period. If the d-axis current is maintained for more than this minimum period, the basic stability of the NLMS adaptive filter operation can be guaranteed as well as the acquisition of samples necessary for parameter identification of the NLMS adaptive filter.

Meanwhile, as shown in FIG. 4, according to the structure of the NLMS adaptive filter constituting the second estimation unit 320, the relationship between the output of the NLMS adaptive filter and the error component is expressed as Equation 7.

는 NLMS 적응 필터의 출력 신호 벡터를,

는 고정자 저항

와 회전자 쇄교 자속

를 추정하기 위한 NLMS 적응 필터의 계수 벡터를,

는 기대 출력 신호 벡터

와 필터의 출력 신호 벡터

간의 순시 오차 벡터를 각각 나타낸다.In equation 7,

is the output signal vector of the NLMS adaptive filter,

is the stator resistance

and rotor flux linkage

The coefficient vector of the NLMS adaptive filter for estimating

is the expected output signal vector

and the output signal vector of the filter

Indicates the instantaneous error vector between the livers, respectively.

) 및 회전자 쇄교 자속 추정값(

)을 나타내는

를 아래의 수학식 8에 따라서 추정한다.The second estimation unit 320 calculates the stator resistance estimate values of Equation 6 and Equation 7 (

) and rotor flux linkage estimates (

) representing

is estimated according to Equation 8 below.

는 스텝 사이즈 행렬을,

는 적응 이득 행렬을,

는 양수 성분으로 이루어진 대각 행렬 형태의 정규화 인자로써, 수학식 8의 두 번째 식의 우변의 역행렬이 특이 행렬의 역행렬 형태가 되는 것을 방지하기 위한 행렬이다. 여기서,

,

및

는

,

및

과 각각 동일한 방식으로 결정될 수 있다.Since the relational expression shown in Equation 8 above is a general formula for an adaptive filter according to the NLMS algorithm, detailed description will be omitted. In Equation 8 above,

is the step size matrix,

is the adaptive gain matrix,

is a normalization factor in the form of a diagonal matrix composed of positive components, and is a matrix to prevent the inverse matrix on the right side of the second equation of Equation 8 from becoming the inverse matrix of a singular matrix. here,

,

and

Is

,

and

and can be determined in the same way, respectively.

와 정규화 인자

를 선정하여야 한다.

와

는 수학식 4에서

과

으로 나타나며, 수학식 8에서

,

로 나타난다. 이때,

가 아래의 수학식으로 정해지는 조건에

가 만족할 경우 제 1 추정부(310) 및 제 2 추정부(320)를 구현하는 NLMS 적응 필터들은 안정적으로 수렴한다.Meanwhile, in order to implement the first estimator 310 and the second estimator 320 of the present invention using the NLMS algorithm, an appropriate adaptation gain is required in Equation 4 and Equation 8.

and normalization factor

must be selected.

and

In Equation 4,

class

It appears as, and in Equation 8,

,

It appears as At this time,

Under the conditions determined by the equation below,

If is satisfied, the NLMS adaptive filters implementing the first estimator 310 and the second estimator 320 converge stably.

가 반드시 수학식 9의 범위내로 선정되어야만 파라미터 추정의 수렴성과 안정성을 보장할 수 있다. 일단, NLMS 적응 필터의 안정성과 수렴성이 보장된 후에는, 수학식 9의 범위 내에서 적응 이득을 조정하고, 정규화 인자를 조정함으로써 각각의 개별적인 구동 시스템에 종속적인 추정 수렴 명세에서 요구하는 과도 응답 성능을 만족 시킬 수 있다.Therefore, the adaptation gain

Must be selected within the range of Equation 9 to ensure convergence and stability of parameter estimation. Once the stability and convergence of the NLMS adaptive filter are guaranteed, the transient response performance required by the estimated convergence specification dependent on each individual drive system is adjusted by adjusting the adaptation gain and the normalization factor within the range of Equation 9. can satisfy.

This fluctuation pattern of the transient response has the characteristic that the transient response becomes faster as the adaptation gain increases or the normalization factor decreases, and conversely, it becomes slower as the adaptation gain decreases or the normalization factor increases.

So far, an apparatus and method for estimating parameters of a SPMSM driving system using an NLMS adaptive filter according to a preferred embodiment of the present invention has been described.

In the case of the prior art, the electrical parameters of the SPMSM driving system were estimated using the Recursive Least Square (RLS) algorithm or the Extended Kalman Filter (EKF) algorithm, but the present invention uses the NLMS algorithm instead of the previously used algorithms such as RLS or EKF. A device for estimating the electrical parameters of the SPMSM driving system was implemented using .

Although the NLMS algorithm has a disadvantage in that the number of calculation iterations required for the filter operation to converge to steady-state performance is higher than that of the RLS algorithm, the computational complexity of one cycle in the case of NLMS is significantly lower than that of RLS, etc. Because the amount of memory is small, two NLMS adaptive filters can be operated at a fast speed equal to the system sampling period, and as a result, convergence speeds such as RLS can be achieved.

In addition, the NLMS algorithm has lower computational complexity and memory requirements compared to RLS, etc., so it is relatively easy to implement while achieving the same convergence speed, and the operating frequency of the parameter estimation device 300 can be adjusted to that of the entire motor drive system 100. It can be applied by setting various settings within the same or lower range as the current control loop operating frequency, which has the fastest operating frequency in the application.

To be more specific, in general, in achieving a similar level of steady-state performance, the NLMS algorithm requires more calculation repetitions than the estimation method based on the RLS algorithm. For this reason, most adaptive filter-based parameter estimation methods are based on the RLS algorithm. However, in terms of computational complexity, the NLMS algorithm shows a lower level than the RLS algorithm, and the amount of memory required for calculation is also significantly lower for the NLMS algorithm than the RLS algorithm.

The online parameter estimation method always involves steady-state control of the motor during parameter estimation calculations. Generally, when controlling the steady state of a motor, the routine that operates at the fastest frequency corresponds to current control. Therefore, if the estimation operation is incorporated into the current control routine, the fastest parameter estimation response can be expected.

However, in the case of an estimation method based on the RLS algorithm, a rapid increase in CPU load ratio cannot be avoided when performing estimation calculations at the same speed as the current control loop due to the high computational complexity and memory amount. In addition, since various user-based programs exist in addition to the current control routine, it becomes inevitable to install a high-performance MCU to drive stable operation.

Regarding this problem, in the prior art for estimating parameters based on the RLS algorithm, the overall estimation process is divided into one operation process for estimating stator inductance and one operation process for estimating stator resistance and rotor flux linkage. there is.

In addition, among the two operation processes, the first operation process that estimates only the stator inductance operates at the same speed as the frequency of the current control program, but the operation process that estimates resistance and magnetic flux at the same time operates at 1/50 times or less than the first operation process. It is operated at a frequency of .

By adopting this method, the CPU load ratio can be drastically reduced, but a sharp increase in the convergence time of resistance and magnetic flux estimates due to the operating frequency of the second operation process cannot be avoided.

On the other hand, in the NLMS-based estimation method according to the present invention, the computational complexity of one cycle is lower than that of the RLS algorithm, so even if the estimation operation is operated at the same frequency as the current control loop, the increase in CPU load rate is lower compared to the RLS-based estimation method.

Therefore, the estimation method according to the present invention is easier to design and implement than the parameter estimation technique using an adaptive filter based on the RLS algorithm, but has the effect of securing the same performance of steady-state error and responsiveness.

Figures 5 and 6 are graphs comparing performance when parameters are estimated using the NLMS adaptive filter of the present invention and when parameters are estimated using the RLS adaptive filter of the prior art.

), 고정자 인덕턴스의 추정값(

), 회전자 쇄교 자속의 추정값(

) 획득하는 동안의 응답을 도시하였다. 이 때, 도 5는 q 축 전류가 1A 에 이르는 부하 상태에서 실험한 것이며, 도 6은 q 축 전류가 3A에 이르는 부하 상태에서 실험한 것이다.The experimental results in Figures 5 and 6 show that from the initial operating state to the convergence state, each adaptive filter produces an estimated value of stator resistance (

), an estimate of the stator inductance (

), an estimate of the rotor flux linkage (

) Responses during acquisition are shown. At this time, Figure 5 shows an experiment under a load where the q-axis current reaches 1A, and Figure 6 shows an experiment under a load where the q-axis current reaches 3A.

Additionally, Figures 5(a) and 6(a) show the mechanical angular velocity of the rotor, Figures 5(b) and 6(b) show the d-q axis current, and Figures 5(c) and 6(c) show ) represents the d-q axis voltage, respectively.

Referring to Figures 5 and 6 (a) to (c), application of d-axis current during a specific cycle under constant load causes a decrease in d-q axis voltage (see time 0.5 to 1 sec), but speed control It can be seen that it has no effect.

Figures 5(d) and 6(d) show the estimation results of the stator inductance, Figures 5(e) and 6(e) show the estimation results of the stator resistance, and Figures 5(f) and 6(f) show the estimation results of the rotor. The estimation results for each linkage flux are shown.

As can be seen from the graphs shown in FIGS. 5 and 6, the parameter estimation method using the NLMS algorithm of the present invention has substantially the same parameter estimation performance as the conventional RLS-based electrical parameter estimation method, but the settling time is different from that of the present invention. It can be seen that this is significantly shorter than the RLS-based parameter estimation method. For reference, the performance indicators for this are listed in Table 1 below.

2 % 이내의 값으로 진입하는 일반적인 제어 이론에서 정의하는 정착시간의 측정값을 나타낸다. Table 1 shows the comparison results of the average value of the estimate and settling time between the estimation device based on the method of the present invention and the estimation device based on the RLS adaptive filter in each load state where the q-axis current is controlled to 1 A, 2 A, and 3 A. The average value of the estimate is the average obtained by obtaining samples for 1 second after sufficient convergence operation is completed, and the settling time is the value of the final convergence value.

It represents the measured value of the settling time defined in general control theory that enters a value within 2%.

The electrical parameter estimation method according to the preferred embodiment of the present invention described so far provides continuous stability in controlling the entire system in cases where there is a change or no change in the electrical constant of the SPMSM driving system, and system parameters Based on this, the control performance of control methods that set the gain of the controller can be improved.

In addition, the parameter estimation method according to a preferred embodiment of the present invention has a faster response speed with less memory and calculation amount and substantially the same steady-state estimation error compared to the conventional estimation technique using an RLS-based adaptive filter, thereby improving the existing method. It is easier to apply to actual systems than the adaptive filter technique.

In addition, the electrical parameter estimation method according to a preferred embodiment of the present invention uses a method of first estimating the stator inductance without applying d-axis current, and then calculating the stator resistance and rotor flux linkage by applying d-axis current when necessary. Two adaptive filters can operate with independent operating frequencies.

So far, the present invention has been examined focusing on its preferred embodiments. A person skilled in the art to which the present invention pertains will understand that the present invention may be implemented in a modified form without departing from the essential characteristics of the present invention. Therefore, the disclosed embodiments should be considered from an illustrative rather than a restrictive perspective. The scope of the present invention is indicated in the claims rather than the foregoing description, and all differences within the equivalent scope should be construed as being included in the present invention.

100: motor drive system
110: Control command generation unit
120: motor driving unit
200: Motor (SPMSM)
300: Electrical parameter estimation device
310: first estimation unit
320: Second estimation unit

Claims (14)

delete a first estimation unit that receives the rotational speed, current, and voltage of the motor from the motor driving system, generates an estimated value of the stator inductance of the motor, and outputs the estimated value of the stator inductance to the motor driving system; and
Receives an estimated value of the stator inductance from the first estimation unit, receives the rotational speed, current, and voltage of the motor from the motor driving system, generates an estimated stator resistance value of the motor and an estimated rotor flux linkage value, and generates an estimated value of the stator resistance and rotor flux of the motor. It includes a second estimation unit output as,
The first estimator and the second estimator are implemented as adaptive filters based on the NLMS (Normalized Least Mean Square) algorithm,
The first estimation unit is
When organizing the continuous-time stator side voltage-current state equation of a surface-mounted permanent magnet synchronous motor (SPMSM) on a dq coordinate system rotating at synchronous speed with the stator inductance as a variable,
The stator inductance is set as the estimation target system of the NLMS algorithm, the coefficients combined with the stator inductance are set as input values to the estimation target system of the NLMS algorithm, and the remaining terms not combined with the stator inductance are set as the output values of the estimation target system. An electrical parameter estimation device for a motor, characterized in that it generates an estimated value of the stator inductance by setting. The method of claim 2, wherein the first estimation unit is
Estimation of electrical parameters of the motor, characterized in that the stator inductance is estimated while the motor driving system sets the d-axis current to 0 and controls the SPMSM (Surface-Mounted Permanent Magnet Synchronous Motor) with the q-axis current as a command value. Device. According to claim 2,
The state equation expressed in Equation 1 below is
[Equation 1]

Transformed into the form of Equation 2 below,
[Equation 2]

Output signal of the system to be estimated

, input signal of the system to be estimated

, the stator inductance (L _s ) of the system to be estimated.

Each is set to,
When the input/output relational expression of the adaptive filter for estimating the estimation target system and the error relational equation of the output of the estimation target system and the adaptive filter are respectively expressed as Equation 3 below,
[Equation 3]

(

is the output signal of the adaptive filter,

Is

Adaptive filter coefficients to estimate ,

is the adaptive filter output and

(indicates the instantaneous error between each)
The first estimation unit is calculated using Equation 4 below:
[Equation 4]

(

is the step size,

is an adaptation benefit,

represents the normalization factor, respectively,

silver

(This is a parameter to prevent division by 0)
A motor electrical parameter estimation device characterized by being implemented as an NLMS adaptive filter expressed as . a first estimation unit that receives the rotational speed, current, and voltage of the motor from the motor driving system, generates an estimated value of the stator inductance of the motor, and outputs the estimated value of the stator inductance to the motor driving system; and
Receives an estimated value of the stator inductance from the first estimation unit, receives the rotational speed, current, and voltage of the motor from the motor driving system, generates an estimated stator resistance value of the motor and an estimated rotor flux linkage value, and generates an estimated value of the stator resistance and rotor flux of the motor. It includes a second estimation unit output as,
The first estimator and the second estimator are implemented as adaptive filters based on the NLMS (Normalized Least Mean Square) algorithm,
The second estimation unit is
When organizing the continuous-time stator side voltage-current state equation of a surface-mounted permanent magnet synchronous motor (SPMSM) on a dq coordinate system rotating at synchronous speed using stator resistance and rotor flux linkage as variables,
The stator resistance and rotor flux linkage are set as the estimation target system of the NLMS algorithm, the coefficients combined with the stator resistance and rotor flux linkage are set as input values to the estimation target system of the NLMS algorithm, and the stator resistance and rotor flux linkage are set as input values for the estimation target system of the NLMS algorithm. An electrical parameter estimation device for a motor, characterized in that it generates estimated values of stator resistance and rotor flux linkage by setting the remaining terms not combined with the flux linkage as the output value of the system to be estimated. The method of claim 5, wherein the second estimation unit is
An electrical parameter estimation device for a motor, wherein the motor driving system estimates stator resistance and rotor flux linkage while injecting a d-axis current less than 0 for a certain period of time while driving the motor in a normal state. According to claim 5,
The state equation expressed in Equation 1 below is
[Equation 1]

Transformed into the form of equation 5 below,
[Equation 5]

According to Equation 5 above,

,

, and

is set as in Equation 6 below,
[Equation 6]

When the input/output relationship of the adaptive filter for estimating the estimation target system and the error relationship of the output of the estimation target system and the adaptive filter are respectively expressed as Equation 7 below,
[Equation 7]

(

is the output signal of the adaptive filter,

is the filter coefficient for estimating the stator resistance and rotor flux linkage,

is the output of the filter and

(indicates the instantaneous error between each)
The second estimator uses Equation 8 below:
[Equation 8]

(

is the step size matrix,

is the adaptive gain matrix,

represents a normalization factor in the form of a diagonal matrix composed of positive components to prevent the step size from being divided into components of 0)
A motor electrical parameter estimation device characterized by being implemented as an NLMS adaptive filter expressed as .
delete A method of estimating electrical parameters of a motor in a parameter estimation device, comprising:
(a) the parameter estimation device receives the rotational speed, current, and voltage of the motor from the motor driving system, generates an estimated value of the stator inductance of the motor, and outputs the estimated value of the stator inductance to the motor driving system; and
(b) Using the estimated value of the stator inductance estimated in step (a) and the rotational speed, current, and voltage of the motor supplied from the motor driving system, generating an estimated stator resistance value and a rotor flux linkage estimated value of the motor. and outputting the output to the motor drive system,
Steps (a) and (b) are performed using an adaptive filter based on the NLMS (Normalized Least Mean Square) algorithm,
The step (a) is
When organizing the continuous-time stator side voltage-current state equation of a surface-mounted permanent magnet synchronous motor (SPMSM) on a dq coordinate system rotating at synchronous speed with the stator inductance as a variable,
The stator inductance is set as the estimation target system of the NLMS algorithm, the coefficients combined with the stator inductance are set as input values to the estimation target system of the NLMS algorithm, and the remaining terms not combined with the stator inductance are set as the output values of the estimation target system. A method for estimating electrical parameters of a motor, characterized in that the estimated value of the stator inductance is generated by setting the stator inductance. The method of claim 9, wherein step (a) is
Estimation of electrical parameters of the motor, characterized in that the stator inductance is estimated while the motor driving system sets the d-axis current to 0 and controls the SPMSM (Surface-Mounted Permanent Magnet Synchronous Motor) with the q-axis current as a command value. method. According to clause 9,
The state equation expressed in Equation 1 below is
[Equation 1]

Transformed into the form of Equation 2 below,
[Equation 2]

Output signal of the system to be estimated

, input signal of the system to be estimated

, the stator inductance (L _s ) of the system to be estimated.

Each is set to,
When the input/output relational expression of the adaptive filter for estimating the estimation target system and the error relational equation of the output of the estimation target system and the adaptive filter are respectively expressed as Equation 3 below,
[Equation 3]

(

is the output signal of the adaptive filter,

Is

Adaptive filter coefficients to estimate ,

is the adaptive filter output and

(indicates the instantaneous error between each)
The step (a) is performed using Equation 4 below:
[Equation 4]

(

is the step size,

is an adaptation benefit,

represents the normalization factor, respectively,

silver

(This is a parameter to prevent division by 0)
A method for estimating electrical parameters of a motor, characterized in that it is performed using an NLMS adaptive filter expressed as . A method of estimating electrical parameters of a motor in a parameter estimation device, comprising:
(a) the parameter estimation device receives the rotational speed, current, and voltage of the motor from the motor driving system, generates an estimated value of the stator inductance of the motor, and outputs the estimated value of the stator inductance to the motor driving system; and
(b) Using the estimated value of the stator inductance estimated in step (a) and the rotational speed, current, and voltage of the motor supplied from the motor driving system, generating an estimated stator resistance value and a rotor flux linkage estimated value of the motor. and outputting the output to the motor drive system,
Steps (a) and (b) are performed using an adaptive filter based on the NLMS (Normalized Least Mean Square) algorithm,
Step (b) above is
When organizing the continuous-time stator side voltage-current state equation of a surface-mounted permanent magnet synchronous motor (SPMSM) on a dq coordinate system rotating at synchronous speed using stator resistance and rotor flux linkage as variables,
The stator resistance and rotor flux linkage are set as the estimation target system of the NLMS algorithm, the coefficients combined with the stator resistance and rotor flux linkage are set as input values to the estimation target system of the NLMS algorithm, and the stator resistance and rotor flux linkage are set as input values for the estimation target system of the NLMS algorithm. A method for estimating electrical parameters of a motor, characterized in that the remaining terms not combined with the magnetic flux linkage are set as the output values of the estimation target system to generate estimated values of the stator resistance and the rotor magnetic flux linkage. The method of claim 12, wherein step (b) is
A method for estimating electrical parameters of a motor, characterized in that stator resistance and rotor flux linkage are estimated while injecting a d-axis current less than 0 for a certain period of time while the motor driving system drives the motor in a normal state. According to claim 12,
The state equation expressed in Equation 1 below is
[Equation 1]

Transformed into the form of equation 5 below,
[Equation 5]

According to Equation 5 above,

,

, and

is set as in Equation 6 below,
[Equation 6]

When the input/output relationship of the adaptive filter for estimating the estimation target system and the error relationship of the output of the estimation target system and the adaptive filter are respectively expressed as Equation 7 below,
[Equation 7]

(

is the output signal of the adaptive filter,

is the filter coefficient for estimating the stator resistance and rotor flux linkage,

is the output of the filter and

(indicates the instantaneous error between each)
Step (b) above is performed using Equation 8 below:
[Equation 8]

(

is the step size matrix,

is the adaptive gain matrix,

represents a normalization factor in the form of a diagonal matrix composed of positive components to prevent the step size from being divided into components of 0)
A method for estimating electrical parameters of a motor, characterized in that it is performed using an NLMS adaptive filter expressed as .

Images