TECHNICAL FIELD OF THE INVENTION

The present invention generally relates to the control of generic electric machines. Particularly, the invention relates to an improved control device and method for controlling an electric machine.

DESCRIPTION OF RELATED ART AND BACKGROUND OF THE INVENTION

Pseudo-linear machine models based on time-varying inductance matrix are theoretically correct only when the media, in which the electromagnetic field exists, can be assumed linear. Exclusively in such a case indeed, the n fluxes linked with each of the n windings are correctly expressible as linear combinations of the n currents in their whole domain. Any attempt to improve such models, aimed at incorporating the non-linear behavior, is theoretically not correct, even if it leads to acceptable quantitative results in some operating conditions. Such improved models can generate heavily mistaken results if one pretends to extend them beyond their usual scope of validity.

In substance, the correct modeling of electric machines, accounting also for magnetic non linearities and/or presence of permanent magnets, demands from the very beginning that the theoretical and procedural approach be radically different from the ones still widely used today, which represented the almost unique practicable way in absence of powerful computers.

DE4115338A1 discloses a control device capable of detecting and/or regulating position or angular velocity of a rotor of an electrical machine by using a Kalman filter.

SUMMARY OF THE INVENTION

It is an object of the present invention to provide a control device and a control method, respectively, for the control of an n-windings electric machine which overcome the limitations disclosed above.

It is a particular object of the invention to provide such device and such method, which are more accurate and precise than prior art devices and methods, and by which improved control of the electric machine can be obtained.

These objects, among others, are according to the present invention, attained by devices and methods as claimed in the appended claims.

According to a first aspect of the invention there is provided a real-time control device for controlling an electric machine with ks windings on a stator and kr windings on a rotor, wherein ks + kr = n and wherein either one of ks and kr may be zero. The control device comprises an input via which control commands are capable of being received in real-time and an output via which control commands to a driver of the electric machine are capable of being output in real-time; machine modeling means provided for modeling the behavior of the electric machine in real time; and decision means operatively connected to the input, the output, and the machine modeling means and provided for (i) determining in real-time the control commands to be output to the driver of the electric machine based on input control commands and results from the modeling of the behavior of the electric machine, and (ii) outputting in real-time the determined control commands at the output. The machine modeling means is provided for modeling in real-time the behavior of the electric machine through at least one functional mapping suited for correlating sets of values of electrical and mechanical quantities and/or sets of values of their total or partial derivatives and/or integral functions with one another, wherein the electrical and mechanical quantities comprise winding currents, winding voltages, magnetic fluxes, mechanical displacements, and/or electromagnetic torques or forces, and the functional mapping is comprised of at least one algorithm and/or mathematical equation based on at least one state function associated with the electromagnetic field inside the electrical machine and/or based on at least one partial derivative of the state function. Hereby an accurate and precise control of the electric machine is enabled, especially in presence of non-linear media. It shall be observed that the machine modeling means does not necessarily require any mathematical transformation on the physical quantities defining the domain of the state function. The modeling may thus be performed without such transformation.

In one embodiment, the machine modeling means is provided with a numerical model of the coenergy or energy associated with the electromagnetic field distribution inside the electric machine, and/or at least one partial derivative thereof. The numerical model is in tabular form, e.g. typically the selected output from FEM (Finite Element Method) simulations of the electric machine. Alternatively, it is obtained from numerical computation of an analytical function.

In another embodiment the machine modeling means is provided as an artificial intelligence model (e.g. based on neural networks and/or fuzzy systems) that describes the state function (e.g. coenergy or energy) associated with the electromagnetic field. In a further embodiment, the control device comprises an input via which the aforementioned measured or estimated electrical and/or mechanical quantities of the electric machine are capable of being received and wherein the machine modeling means is provided to identify the coenergy or energy associated with the electromagnetic field distribution inside the electric machine, and/or a partial derivative thereof, based on the measured or estimated quantities of the electric machine. Typically, the measured or estimated quantities of the electric machine comprise at least one current in at least one of the n windings and at least one voltage across at least one of the n windings preferably as measured or estimated while the rotor of the electric machine is stationary in at least one selected angular position.

Other measurable quantities which can be used in the identification of the coenergy or energy associated with the electromagnetic field distribution inside the electric machine, and/or the partial derivative thereof, may comprise at least one electrical resistance of at least one of the n windings, the mechanical torque at the shaft of the electric machine, and/or the total first order derivatives of at least one of the winding currents.

According to a second aspect of the invention there is provided a control method for controlling an electric machine with ks windings on a stator and kr windings on a rotor, wherein ks + kr = n and wherein either one of ks and kr may be zero. According to the method input control commands are received in real-time; the behavior of the electric machine is modeled in real-time; control commands to be output to a driver of the electric machine are determined in real-time based on the input control commands and results from the modeling of the behavior of the electric machine; and the determined control commands are output to the electric machine in real-time. The modeling of the behavior of the electrical machine is made through at least one functional mapping suited for correlating sets of values of electrical and mechanical quantities and/or sets of values of their total or partial derivatives and/or integral functions with one another, wherein the electrical and mechanical quantities comprise winding currents, winding voltages, magnetic fluxes, mechanical displacements, and/or electromagnetic torques or forces, and the functional mapping is comprised of at least one algorithm and/or mathematical equation based on at least one state function associated with the electromagnetic field inside the electrical machine and/or based on at least one partial derivative of the state function. Further characteristics of the invention and advantages thereof will be evident from the following detailed description of embodiments of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS

The present invention will become more fully understood from the detailed description of embodiments of the present invention given herein below and the accompanying Figs. 1-5, which are given by way of illustration only, and thus are not limitative of the present invention.

Fig. 1 displays schematically a control device for controlling an n-windings electric machine according to an embodiment of the invention.

Fig. 2 displays schematically a driver and an n-windings electric machine according to an embodiment of the invention.

Fig. 3 displays schematically a sensing winding embedded in a coil winding (upper left portion) and in radial cross section an electric machine with a rotor and a stator, both having coil windings and sensing windings embedded therein.

Figs. 4a-b display schematically two arrangements for measuring the torque of the electric machine at standstill, the arrangements being capable of fine controlling the shaft angle. The right hand portion of the respective Figure displays in a top view the arrangement mounted in the electric machine whereas the left hand portion of the respective Figure is a cross section taken along line A - A of the top view.

Fig. 5 displays schematically an arrangement for measuring total current derivatives of the currents in the power connections from a driver to the electric machine. The arrangement features sensing coils in series with, or arranged around, the power connections to the electric machine.

DETAILED DESCRIPTION OF EMBODIMENTS

In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular techniques and applications in order to provide a thorough understanding of the present invention. However, it will be apparent to one skilled in the art that the present invention may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known techniques are omitted so as not to obscure the description of the present invention with unnecessary details.

A real-time controller or control device 11 for controlling an n-windings electric machine according to an embodiment of the invention is displayed schematically in Fig. 1. The electric machine has ks windings on the stator and kr windings on the rotor wherein ks + kr = n and wherein either one of ks and kr may be zero. The control device 11 comprises an input 12 via which control commands are capable of being received, optionally an input 13 via which measured or estimated quantities of the electric machine are capable of being received, and an output 14 via which control commands to a driver of the electric machine are capable of being output. The control device further comprises machine modeling means 15 and decision means 16, preferably provided as a microcomputer device provided with suitable software.

Fig. 2 displays schematically a driver 21, e.g. a variable speed drive (VSD) system, in which the control device 11 may be arranged. The driver 21 is operatively connected to an n-windings electric machine 22 for driving the same.

The machine modeling means 15 of the control device 11 is provided for modeling the behavior of the electric machine 22 by using a model based on a state function associated with the electromagnetic field distribution inside the electric machine 22, and/or based on at least one of its partial derivatives. The state function may be the energy, but is more comfortably the coenergy associated with the electromagnetic field distribution inside the electric machine 22, or partial derivatives thereof. The decision means 16 is operatively connected to the inputs 12, 13, to the output 14, and to the machine modeling means 15 and is provided for determining the control commands, and/or on measured or estimated quantities 13, to be output to the electric machine 22 at the output 14 based on input control commands and results from the modeling of the behavior of the electric machine as performed by the machine modeling means 15. Subsequently, the decision means outputs determined control commands at the output 14. The modeling means may be connected directly to the inputs 12, 13 and the output 14.

The coenergy-based model of the electric machine 22, characterized by one moving part (the "rotor") rotating around an axis can be described by the following equation system:

wherein the first equation of the system is provided in a plurality, expressed by the natural index j, corresponding to the number n of windings. The index j ranges from 1 to n, and the symbols present in the equations are identified as follows:

Scalar value of the angular position of the rotor

Scalar value of the rotational speed of the rotor around its

axis assumed stationary in an inertial reference frame

Vector of the n winding currents.

Voltage and current of the f ^h winding (a two-terminal network) / e {l..n} . They are referred to the users' convention

with the current reference direction entering the winding terminal.

Equivalent total moment of inertia of the rotating masses. It

is referred to the rotor axis of rotation. (C)

Resistance of the f ^h winding (C). / e {l..n}.

Coenergy of the electromagnetic field in the overall machine space, expressed as state function of the winding

currents and angular position of the rotor

Voltage and current of the f ^h winding (a two-terminal network) j€ {l..n}. They are referred to the users' convention

with the current reference direction entering the winding terminal.

Scalar value of the torque exerted by the mechanical system coupled to the electrical machine. It is referred to the

rotor axis of rotation

By trivial manipulation the equation system above can be rewritten in the form shown below: r d ² Wco( 7(f) , θ„, (0 ) ] di _k (t) \ d ² Wco( t) , e _m (t) )

diidik dt j dijd0 _m

d ' _m (t) _ dWco( l(t) , θ, _η ( ή de _m (t

wherein the first equation of the system is provided in a plurality, expressed by the natural index j, corresponding to the number n of windings and the symbols of the equations are identified as above.

By observing the above equations it can be noted that once suitable partial derivatives of the coenergy are known, the model for control purposes is completely defined and computable. The iron losses are not comprised, but can be modeled separately.

The machine modeling means can be provided with the coenergy partial derivatives either numerically in tabular form or in a discretized form of an analytical function or an artificial intelligence descriptor, preferably consisting of a set of artificial neural networks and/or fuzzy inferences.

. The partial derivatives can also by definition be calculated from the coenergy state function, if this is known with sufficient accuracy for numerical differentiation. If the coenergy is known analytically, the partial derivatives are derived by analytical differentiation. If the coenergy is known in tabular form, a sufficient number of points are needed to limit the numerical error affecting the numerical differentiation techniques that can be employed. It should be highlighted that the computation of the coenergy or the energy is a standard feature in almost all FEM simulators today. For a specific machine design it is therefore procedurally straightforward to obtain the coenergy function in a tabular form for a chosen set of currents and angle values.

The models of any electromechanical converter which are based on state functions (energy or coenergy) associated with the electromagnetic field distribution have always been recognized as the only correct ones, especially in presence of non-linear media. Their adoption has always been impractical in the past because of the lack of computational power, both for what concerns the machine design and machine control aspects, as well as for what it concerns the instrumentation, which was incapable of intense and complex data acquisition and post-processing. Conversely nowadays, the availability of inexpensive fast computers, even inside the real-time controllers and the instrumentation, enables the machine analysis and control to the aforementioned correct fundamental models based on state functions, most comfortably the coenergy.

The modeling approach has the advantage of stemming from the unified vision of the electromechanical conversion, since; ultimately, only one electric machine exists. The approach is valid for all different types of electric machines. It is therefore suitable for a standardization of treatments, both in design/simulation and control, with consequent cost reductions brought by the possible uniformation of analysis and control platforms.

If the coenergy function is not known analytically or through simulations, the machine modeling means 15 can be provided to identify the coenergy associated with the electromagnetic field distribution inside the electric machine 22 based on the measured or estimated quantities of the electric machine 22 as received via the input 13 of the control device 11.

It shall also be appreciated that even if the coenergy function is known analytically or through simulations this estimated coenergy function may differ from how the electric machine actually behaves. In such instance, measured or estimated quantities of the electric machine 22 as received via the input 13 of the control device 11 may be used to enhance the already existing coenergy model of the electric machine.

Any method of such machine identification, especially if based on state functions, strongly benefits from a good knowledge of the fluxes linked with the machine windings. Such fluxes in principle can either be directly measured with sensors sensible to the magnetic field or indirectly obtained by integration of the voltages induced in them. This latter approach is still the technologically simpler nowadays, but it is also prone to errors that the following proposed aspect of the invention is meant to reduce.

If all windings are accessible the coenergy can be determined based on currents in the n windings and voltages across the n windings as measured while the rotor of the electric machine is stationary in a selected angular position, see Equation 3.2 in the attached Appendix below.

A main obstacle in good flux determination through time integration of voltages is the necessary knowledge of the winding resistances and their variations which must be continuously tracked for maintaining an acceptable accuracy. In order to overcome this limitation, especially for more valuable high-power machines and processes, it is proposed the addition of small "sensing" windings in the electric machine; windings disposed in such a way that each of them follows as much as possible the geometrical path of the normal "power" winding to which it is associated. If such a condition is satisfied the sensing winding is subject to a field distribution almost identical to the one of the corresponding power winding. As a consequence the voltage induced in the sensing winding, and the flux linked with it, are proportional, with very good approximation, to the voltage induced in the power winding and its linked flux, respectively. The factor of proportionality is clearly the ratio among the number of turns in the sensing and power windings. If the field is sufficiently intense the sensing winding could be realized very simply by one single turn of wire disposed along the path of the associated power winding.

Although different arrangements of this principle are possible, a possible one is to embed a sensing coil - made of very thin wire with proper voltage insulation level - inside each coil composing the power winding as being illustrated in Fig. 3. The sensing wire 32 can be embedded during the wrapping of the coil insulation layers so that each coil 31 receives two additional terminals which belong to the sensing winding 32. As being illustrated in the upper left portion of Fig. 3. Afterwards, when the coils are connected properly to compose a winding, their respective sensing wires are connected in the same order thereby realizing straightforwardly the sensing winding, which develops itself all along the "curve" of the power winding, as can be seen in the lower right portion of Fig. 3, in which the coil windings 31 and the sensing windings 32 are visible both on the rotor and on the stator of the electric machine.

All terminals of the sensing windings are then concentrated into a separate terminal box of the electric machine, where the voltages can be measured and sent to the drive controller, or even amplified and integrated locally if necessary or requested by the customer. Clearly these voltages truly reflect the total time derivative of the magnetic fluxes linked with the power windings only, without any undesired component due to the resistive drop that is therefore rendered completely uninfluenced. There is no longer any need of tracking or even knowing the resistances of the power windings.

An electric machine with these sensing windings can be offered to the customer as an enhanced version for more performing control and/or diagnostic. The additional manufacturing cost is limited but the additional value for the customer and drive manufacturer can be significant. As mentioned previously, the integration of the voltages could be performed even locally, in the terminal box of the sensing windings, similarly to what is done by the integrating amplifiers present in the current sensors based on Rogowski coils. Electric machines can therefore be manufactured that already provide either the signals proportional to the important fluxes above a certain speed (i.e. the already integrated sensed voltages), or the sensed voltages only, leaving in this case the integration to a separate purchasable set of amplifiers designed for this purpose.

It can be observed that it is not necessary to equip all power windings with an associated sensing winding. One can limit itself to the minimum number of sensing windings which are necessary for the machine identification or sensorless control, as described in detail in the Appendix below. In order to adhere to a common commissioning requirement, it has been chosen to identify the coenergy-based model of the electric machine, for what is possible, keeping the rotor locked at standstill. This locking is realized by suitable mechanical means.

If all windings of the machine have observable currents, the coenergy can be determined with measurements at the electrical terminals only through the Equation 3.2 of the attached Appendix below. The details are described in section 3.1 and the steps in 3.1.1. More accurate measurement of the coenergy or identification of the input-output mapping caused by it - sufficient for electric machines with non observable windings - demand the measurement of the electromagnetic torque at standstill and at different shaft angles imposed during the procedure.

Since the torque is measured at standstill the use of an expensive torsiometer - aimed at measuring the torque during rotation - is not necessary. A more accurate measurement technique can use load cells (usually more accurate than a torsiometer) and a sufficiently stiff mechanical arrangement that connects them to the machine base while allowing the imposition of the desired angle.

One arrangement, illustrated in Fig. 4a, comprises two load cells 41 connected to a yoke 42 arranged around the shaft 44 in a friction-based coupling which is tightened by means of bolts 43 once the desired angle is imposed. This solution is suitable for shafts 44 already coupled to the load machinery, without possibility of being disconnected. In the other end the load cells 41 are connected to a fixed support connected to the base of the electric machine.

An alternative arrangement uses a yoke or leverage tightened to the machine shaft as above, but held steady by one (or two) hydraulic ("oleodynamic") pistons instead of load cells. The force can be transmitted by the hydraulic oil in the pressure pipes to a remote and more comfortable location where another receiving piston acts on one load cell. This is a much more flexible solution for the logistic and mounting, but care must be put in assuring that the deformation of the pipes and the compressibility of the hydraulic oil does not introduce excessive systematic errors in the measurement.

A yet alternative arrangement, illustrated in Fig. 4b, is coupled to the flange of one free side of the machine shaft 44, when available. Such a solution is the most comfortable from the operational point of view. It uses a worm-thread coupling whose pinion 46 is connected to the shaft 44. The pinion 46 has a helicoidal profile 47 which is brought into engagement with a thread 48, thereby converting the torque into an axial load which can be sensed by a load cell 45 connected axially between an end of the thread 48 and a fixed support of the electric machine. The pinion 46 could possibly be machined onto the machine shaft 44 itself, if the nominal torque to be withstood is not excessive. Because of the non reciprocity of the worm-thread coupling (provided that the thread angle is below a certain value), the arrangement is self-blocking for movements originating inside the machine. Conversely, the shaft can be easily and very precisely rotated by acting on the thread, either automatically via a servomotor 49 that allows also an automation of the identification procedure, or manually via a handle 50.

All apparatuses are removed at the end of the identification procedure.

The coenergy, like the energy, is a state function of the electromechanical system; therefore its value depends only on the point defined by its arguments, and not on the trajectory followed to reach that point. One can exploit this property to determine the coenergy more accurately through the following procedural choices, described more extensively in the attached Appendix:

- The electrical sources (VSDs or others) are connected to the machine winding and controlled so that they behave as current sources with DC waveforms in steady state.

- The desired set of winding current values (one "point") is reached by varying linearly one current at the time, never two simultaneously. The other currents are kept at their preceding constant value while the selected one is varied. During these controlled transients the voltages and currents in all windings are recorded. With such recordings the coenergy value is updated through the Equation 3.2 of the attached Appendix.

- The measured values of the winding resistances are updated only when all currents are constants because in such time intervals all winding voltages are composed by the resistive components only. Before performing the ratio between the measured voltages and currents, the measurements are averaged for a sufficiently long time in order to eliminate the systematic errors caused by the noise and unavoidable ripple presence, both characterized by zero average value.

During the linear variation of one winding current only - while the others are kept constant - an even more direct determination of important partial derivatives of the coenergy can be performed. Such derivatives would constitute the elements of the matrix of inductances in case the machine was perfectly magnetically linear. This possibility encourages the direct measurement, in an analogical way, of the current derivatives. It is possible to do this with limited cost increase by inserting small calibrated inductances in series to each line connecting the drive to the machine as being illustrated in Fig. 5. The measurement of the voltage across such inductances is clearly proportional to the current derivative. An alternative, also shown in Fig. 5, employs sensing coils 51 (without integrating amplifier), e.g. Rogowski coils, mounted around the conductors. Rogowski coils without amplifiers are easily available as standard components ready to be purchased.

This latter solution has also the great advantage of being applicable to already existing converters - potentially under refurbishing or revamping - without the need of modifying their construction, as it would be the case when series inductances were used. Such converter might receive a new control device which incorporates the sensing of the Rogowski voltages, thereby increasing their capabilities and customer value at little additional cost in the revamping process.

Also the sensorless determination of the speed is improved by measuring the current derivatives directly. This is an additional reason for providing the VSD system with this capability.

It shall be noted that the present invention is not limited to a control device, but encompasses as well a driver, such as the driver 21 of Fig. 2, for an n- windings electric machine comprising the inventive control device.

Yet further, the invention also encompasses an n-windings electric machine 22, such as the electric machine 22 of Fig. 2, provided with a driver comprising the inventive control device.

Still further, the invention encompasses a control method for operating an inventive control device. Such method may comprise any of the method steps, procedure steps, or process steps as disclosed herein.

It will be obvious that the invention may be varied in a plurality of ways. Such variations are not to be regarded as a departure from the scope of the invention. All such modifications as would be obvious to one skilled in the art are intended to be included within the scope of the appended claims. Particularly, it shall be appreciated that the various features and limitations depicted herein can be used separately or can be combined in each possible manner to obtain different embodiments of the invention. Attached hereto is an Appendix discussing in detail the concept of the coenergy model and the identification thereof based on experimental measurements. The Appendix is an integral part of the Detailed description of embodiments.