下載/瀏覽

Download Report

Transcript 下載/瀏覽

Improved Speed Estimation in Sensorless
PM Brushless AC Drives
Authors: J.X. Shen, IEEE Member
Z.Q. Zhu, IEEE Senior Member
David Howe
IEEE Transactions on Industry Applications,
Vol.38, No.4, August 2002
Student: Sergiu Berinde
Outline

Abstract

Introduction

Flux-Observer-Based Sensorless Control

Speed Estimation From Differential of Rotor Position

Filtering of Observed Flux Vector

Average Speed Estimation

Improved Method of Speed Estimation

Conclusions
Abstract




The application of flux-observer-based sensorless control to PM brushless AC
motors is described
Methods in which the speed is derived from the differential of the rotor
position and from the ratio of the electromotive force to excitation flux linkage
are assessed.
An improved method combining the best features of both previous methods is
proposed.
The performance of the methods is verified experimentally.
Introduction

Two major categories of sensorless control techniques :






1. speed is estimated from an observer and rotor position is obtained by
integration
2. rotor position is estimated from an observer and speed is calculated by
differentiation
In general, derivation of speed from rotor position => significant noise
Estimation of average speed is accurate under steady-state conditions, but
not fast enough to give good dynamic response.
Estimation of speed from the induced EMF and excitation flux linkage => fast
dynamic response, but low accuracy.
Improved method of speed estimation would be needed.
Introduction


A TMS320C31 DSP-based drive, which supplies a two-pole surface mounted
PM BLAC motor is used in the investigation.
An encoder is used to measure the rotor position and deduce the speed from
the differential of the position. These values are compared with the ones
obtained from the sensorless techniques.
Flux-Observer-Based Sensorless Control

The phasor diagram of a BLAC is shown.
•
- excitation flux linkage due to permanent
magnets, in phase with the rotor d axis
•
- resultant stator winding flux linkage
•
- angle of rotor, we want to know it
• We know that
• then,
• We also know that
•
is expressed by
,
• Therefore, the rotor position is obtained as
Flux-Observer-Based Sensorless Control

The root locus of the observed flux vector
is shown.
• displaced due to integration in
• solution: apply HPF to the variables to be integrated,
equivalent to replacing the integrator by a LPF
• after applying the HPF, the circular locus of
remains stable
Flux-Observer-Based Sensorless Control

Estimated rotor position and actual position are compared.

Expressed in terms of encoder resolution of 4000 pulses / revolution

Maximum error is 50 pulses, or 4.5˚ electrical => sufficiently accurate for
vector control of most drives.
Speed Estimation from Differential of
Rotor Position

The rotor angular velocity

Since, in general, errors will exist in

is given by
and
, then
Since
is very short,
is comparable with
=> can
cause significant error in the estimated speed and errors become greater at
lower speed.
Speed Estimation from Differential of
Rotor Position


In order to demonstrate the limitation of this method => speed change every
2 sec, with fuzzy algorithm for speed control
Estimated speed contains significant ripple => inappropriate for feedback in
sensorless drive system
Filtering of Observed Flux Vector


The ripple in the estimated speed is caused by position errors, which in turn
arise from noise in the observed flux-linkage vector locus => try to apply LPF
to the observed flux vector
The flux vector estimation is improved, but a steady-state error of 100
pulses, or 9˚ electrical still exists in the estimated rotor position
Filtering of Observed Flux Vector

Cause of error => phase shift introduced by the LPF
Average Speed Estimation



Since a flux filter is not always effective in reducing the speed error => apply
another LPF to filter the speed
=> obtain average speed
When applying the speed filter, effects of the flux filter are reduced even at
high speed
is in close agreement with
under steady-state conditions
Average Speed Estimation



However, the speed filter introduces a noticeable time delay in the estimated
speed
Does not introduce phase shift, unlike the flux filter, however because of the
introduced time delay, it may compromise the dynamic response of the drive
Accurate for steady-state conditions, but slow
Speed Estimation From Induced EMF and
Excitation Flux Linkage




If
is known a priori and the amplitude
of the EMF is calculated from
the measured current and voltage, then in the d-q rotor reference frame
We can express
as
But
,
are sensitive to variations in temperature and the differential
operation (
) can cause a significant error in the estimated speed => this
method is inherently inaccurate
However, by eliminating the differential operation => lower accuracy, but fast
response
Speed Estimation From Induced EMF and
Excitation Flux Linkage


Speed change to verify the fast response of the method
The most important feature of the method => response is fast, although
accuracy of the estimation is not good
Improved Method of Speed Estimation

Two conclusions can be drawn:


When using
When using
, a time delay exists between estimated and actual speed
, a magnitude error exists between estimated and actual speed

However, the two methods can be combined to improve the estimation

If s -> 0 , then

We can write

->
, and if s -> ∞ , then
->
as
The method compensates
of
,
by adding the output of the high-pass filtering
Improved Method of Speed Estimation
Conclusions




Various methods for estimating the speed of a sensorless BLAC motor drive
have been implemented and compared.
Estimation of the speed from the differential of rotor position -> accurate
under steady-state but poor dynamic response
Estimation of the speed from the EMF and excitation flux linkage -> not
accurate, but fast response
The proposed method of speed estimation has been shown to yield improved
performance and is considered suitable for closed-loop speed control