METHOD FOR STARTING AND STOPPING AN ASYNCHRONOUS ENGINE

Description

[0001] The invention relates to a method for starting and stopping an asynchronous motor. The invention also relates to a soft starter. Likewise, the invention relates to a computer program product with which the method can be carried out.

[0002] Soft starters, also known as soft start devices, are devices that utilize power electronic control via thyristors. These thyristors, when used with conventional control methods, reduce the mains voltage to improve the starting of asynchronous motors (also called induction motors). Soft starters are used in various applications, including pumps and fans in industrial settings. They are particularly useful when the starting currents of asynchronous motors, which are typically 6 to 10 times the rated current, need to be reduced, or when the starting torque of a starting asynchronous motor needs to be lowered, as this torque is usually significantly higher than the rated torque, at least in certain sections. Soft starters typically remain active until the motor reaches its rated speed, thus completing the starting process.In industrial devices, once the start-up process is complete, so-called bypass contacts are often activated. These take over the current flow and connect the ASM directly to the mains. Conducting the current via the bypass contacts has the significant advantage of eliminating losses in the thyristors.

[0003] FIG 1 Figure 1 shows a typical soft starter topology with a soft starter 1 connected between a three-phase electrical network 5 and an asynchronous motor 4 for starting the motor. The soft starter 1 has one antiparallel pair of thyristors 2 and one bypass contact 3 per phase a, b, c. The current ia, ib, ic is measured on the motor side in all phases a, b, c. Additionally, the soft starter 1 uses line-to-line measurements uA,B, uB,C, uC,A on the network side to calculate the network voltages uA, uB, uC, and line-to-line measurements ua, b, ub, c, uc, a on the motor side to calculate the motor voltages ua, ub, c, uc, a. Similar topologies, in which only two of the three network phases can be switched in the soft starter, can be found, for example, in... B. in the Siemens device manual soft starter SIRIUS 3RW50, Siemens AG, Amberg, 09 / 2019, A5E35628455001A / RS-AA / 001, chapter A.1.

[0004] Although the range of functions of soft starters has been continuously expanded in recent decades, e.g. through voltage ramping, current-limited operation, torque-controlled start-up and speed-controlled start-up, the structure for generating control pulses has remained essentially the same and has not changed in the FIG 2 The structure shown, which essentially comprises three blocks, is explained below: In the first block of the soft starter, the controller 21, a control structure generates the setpoint of the ignition angle α, also called the control or trigger angle, from measured values ​​24 such as stator current I1,RMS, stator voltage U1,RMS, and / or rotor speed n of the motor. The ignition angle α can be influenced by considering various parameters 25, such as the motor torque, the maximum set current, or a speed ramp, so that the behavior is optimized for the application. The ignition angle α defines the RMS value of the voltages applied to the individual phases a, b, c of the motor (RMS = Root Mean Square). A simple way to generate an ignition angle α is a voltage ramp, independent of the load response; for this purpose, the ignition angle α is simply increased as a function of time.It is necessary to start with a sufficiently large initial value of the ignition angle α so that the initial motor torque MM is already greater than the load torque ML; otherwise, the motor current would not accelerate the rotor and would only generate high losses at standstill until the motor torque is high enough to accelerate the rotor.

[0005] A second block 22 of the soft starter serves to generate trigger signals 27 for triggering thyristor ignitions, whereby the trigger signals 27 are generated on the basis of the measured mains voltage 26.

[0006] FIG 3 The figure shows the relationship between the voltage u across the thyristor, the current i flowing through the thyristor, and the firing angle α as a function of time t. The conventional soft-start algorithm uses the principle of voltage reduction to start ASMs. From a firing angle α, a reduction in the mains voltage is achieved by continuously firing the thyristors at the same firing angle. This reduction in the mains voltage leads approximately to a proportional reduction in the starting current: I 1 , RMS ∼ U 1 , RMS

[0007] The stator current I1,RMS of the motor, or simply motor current, decreases proportionally to the applied motor voltage U1,RMS. Since the motor torque MM depends approximately quadratically on the motor current I1,RMS, a significant reduction in motor torque MM also occurs, even with smaller reductions in the motor voltage U1,RMS. M M ∼ I 1 , RMS 2 ∼ U 1 , RMS 2

[0008] A third block 23 of the soft starter comprises the thyristors and ignition devices, e.g. an ignition signal unit; it uses the ignition signals received from the second block 22 to generate the output voltage.

[0009] A conventional soft-start algorithm uses steady-state values ​​for control, e.g., of the motor current I1,RMS. All considerations are based on the assumption that the system is more or less in a steady state and that the slip changes over time.

[0010] The losses incurred when starting an asynchronous motor with a soft starter are problematic. Firstly, the thyristors themselves generate losses within the device. This necessitates adequate cooling, particularly for the thyristors. Secondly, increased losses occur within the motor during startup. These are significantly higher than the typical motor losses during rated operation. These two factors limit the number of consecutive starts that can be performed.

[0011] CHOI WOO JIN ET AL: "A novel MPC-SVM strategy for direct torque flux control of an induction motor drive system using a matrix converter", 2014 IEEE INTERNATIONAL CONFERENCE ON INDUSTRIAL TECHNOLOGY (ICIT), IEEE, February 26, 2014 (2014-02-26), pages 181-186, XP032639917, 001: 10.1109 / ICIT.2014.6894935, describes a method for torque and flux control of an induction motor using model predictive control space vector modulation (MPC-SVM). In this process, nine bidirectional switching elements of a 3-phase matrix converter (MC) are switched to electrically conductive (opened) or blocking (closed) at suitable times in order to generate desired voltages in the three phases.

[0012] EP 1 240 712 B1 (SIEMENS AG [DE]) 24 September 2003 (2003-09-24) describes a method for finding suitable ignition timings of a three-phase controller for operating an asynchronous motor. This method generates operation of an asynchronous machine with arbitrary fundamental frequencies solely by means of a three-phase controller with three or two pairs of antiparallel thyristors. Another predictive three-phase controller is disclosed in EP 2 571 157 A1.

[0013] The invention is therefore based on the objective of reducing the losses occurring during the start-up of an asynchronous motor connected to a soft starter.

[0014] The problem is solved by a method according to claim 1. The problem is also solved by a soft starter according to claim 9. The problem is also solved by a computer program product according to claim 10.

[0015] The method according to the invention serves to start and stop an asynchronous motor using a soft starter. The soft starter is connected in the electrical supply line between an electrical network supplying the asynchronous motor with electrical energy and the asynchronous motor itself. The soft starter has thyristors connected in antiparallel in the switchable phases, preferably in two or three phases of the supply line, so that the soft starter can operate as a current controller. Starting an asynchronous motor is also referred to as ramp-up or starting; this term describes the phase between the motor being at standstill and the point at which the motor reaches its rated speed. Stopping an asynchronous motor is the opposite process to starting: this term describes the phase between the point at which the motor last rotates at its rated speed and the motor being at standstill.In both phases, the motor accelerates: positive acceleration (increase in rotational speed) during starting and negative acceleration (decrease in rotational speed) during stopping. The method according to the invention is suitable for both processes, starting and stopping.

[0016] The procedure has a first step in which possibilities for triggering one or more thyristors of the soft starter are determined for a future calculation point in time. This future calculation point can also be a period or time interval during which triggering one or more thyristors of the soft starter can occur. Possibilities for triggering one or more thyristors of the soft starter are also referred to as triggering possibilities or triggering options. The thyristors used in the soft starter have the property that they can be switched on, i.e., triggered, but not switched off again. They only return to the blocking state when the current crosses zero. This means that a successful triggering event typically sets the current for a period of several milliseconds.Thus, the number of ignition possibilities can be limited if, at the time of calculation, for example, two phases are still conducting current that is still flowing in the system due to an earlier ignition. In a 3-phase network topology, this reduces the number of ignition possibilities to one, since only the third phase can be "ignited".

[0017] The method includes a further step in which the electrical and / or mechanical engine behavior is predicted for the identified ignition possibilities if one or more thyristors of the soft starter were to be triggered. For the ignition possibilities derived from the topology, the resulting behavior is predicted using an engine model. For this purpose, the connected ASM (automatic starter motor) can be modeled using a transient engine model.

[0018] The method includes a further step in which, based on the predicted electrical and / or mechanical engine behavior, a decision is made as to whether and which ignition option should be selected. It is possible that, taking into account the predicted engine behavior and decision criteria within the method, it is decided that none of the identified ignition options should be selected. In this case, no ignition occurs at the calculation time for which no ignition options were selected. Instead, the ignition options can be determined for a later calculation time, preferably for an immediately following calculation time; that is, a repeat of the method, starting from the first step, can follow.

[0019] The process includes a further step in which, if a decision has been made to use an ignition option, one or more ignition signals are generated for the soft starter. This step can then be followed by a repeat of the process, starting from the first step.

[0020] As explained in the introduction, the control mechanism, particularly the pulse generation, in soft starters has changed little in recent decades. This is where the invention comes in: Instead of using the existing structure to generate ignition signals, a new structure is employed that does not operate via ignition angle and does not rely on continuous control of the thyristors.

[0021] The new method employs a model-predictive approach to generate ignition signals for soft starters: first, all possible ignition scenarios in the next time step, the so-called ignition possibilities, are calculated, and then a decision is made for each ignition possibility as to whether it can be used or not. Thus, instead of relying on continuous thyristor ignition, average values, and the assumption of steady states, as in conventional methods, the invention uses a model to predict the transient electrical and / or mechanical motor behavior and makes a decision about ignition based on this prediction. The considerations of the new method are situational and do not assume average values ​​or steady-state behavior. The process is then repeated, either directly or at fixed time intervals.This prediction is preferably only performed for the immediately pending ignition opportunities.

[0022] The basic idea is to use an engine model to predict the electrical and / or mechanical behavior of the engine for all ignition possibilities for at least the next time step. To this end, all state variables of the system are measured and calculated for initialization. Grid voltages uA, uB, uC are measured to calculate the grid angle φgrid and the grid voltage amplitude Ugrid. The engine currents ia, ib, ic and engine voltages ua, ub, uc are also measured for a model-based calculation of the line-linked magnetic flux. Ψ → 2 K The rotor's magnetic flux, hereinafter also referred to as rotor-magnetic flux linkage or simply rotor flux, is used. Together with the measured mechanical rotor speed n and the grid frequency fgrid, all system variables are defined.

[0023] This information is used to calculate all ignition combinations possible with the engine model in the next time step, i.e., the so-called ignition possibilities. In the case of a 3-phase network topology, the result is up to four calculated torque and current variants over time: 1.-3. two conducting phases (a and b, or b and c, or c and a) 4. three conducting phases (a, b and c).

[0024] The calculated results can be checked against various criteria to see if limit values ​​have been exceeded. If a usable ignition opportunity exists, the algorithm generates ignition signals to implement the calculated result in reality.

[0025] The invention has the advantage of reducing losses in the soft starter and the induction motor during the starting process. This makes it possible to perform more starts per unit of time, reduce cooling requirements, and thus build smaller soft starters and save energy when an application is started frequently.

[0026] There are two essential steps in the new procedure: prediction and decision-making.

[0027] The method is based on prediction. The thyristors used in the soft starter have the property that they can be switched on, i.e., triggered, but not switched off again. They only return to the blocking state when the current crosses zero. This means that a successful ignition typically sets the current for a period of several milliseconds. In the case of a 50 Hz AC voltage, a zero crossing of the voltage occurs every 10 ms. Another influencing factor is the combination of engine phases to be ignited. A basic distinction can be made between two-phase and three-phase ignitions. In the case of the FIG 1 The soft starter topology shown thus results in four ignition possibilities: 1. Ignition of phases a and b, 2. Ignition of phases b and c, 3. Ignition of phases c and a, 4. Ignition of phases a, b and c.

[0028] The procedure described here is not applicable to the one described in FIG 1 The topology shown is limited. It only provides a good basis for explanation. Other topologies, including those for two-phase soft starters, can also be operated using the methodology described here, although some impact on performance is to be expected. Other interconnections of power electronic switches would also be possible, e.g., a root-3 circuit or a thyristor reversing circuit. Ultimately, the only consideration here is which thyristors can be triggered together, i.e., which firing scenarios occur. These can then be predicted and evaluated using the described procedure. The methodology remains the same.

[0029] The number of ignition possibilities can be limited if, for example, two phases are still conducting current at the time of calculation, current which is still flowing in the system due to an earlier ignition. In the topology shown here, this reduces the number of ignition possibilities to one, since only the third phase can be "ignited" additionally.

[0030] For the ignition possibilities derived from the topology, a model is used to predict the resulting behavior. For this purpose, the connected motor can be modeled using a transient motor model.

[0031] Using the equation systems presented above, the electrical and / or mechanical behavior of the motor can be predicted. However, the model's states must be initialized before further calculations can be performed. This means the simulation model must be set to the current state from which the prediction begins. For this, the following parameters must be determined: The stator current can be easily measured in the soft starter. The model is then initialized with the current value.

[0032] Unlike stator current, rotor flux cannot be easily measured. Measuring the actual flux in the rotor is almost never implemented in industrial practice due to the high costs involved. The most common solution is the use of a flux monitor or flux model. Using the measured motor currents and / or voltages, the flux in the motor can be reconstructed. Numerous methods exist for this purpose.

[0033] Besides the rotor flux, the motor speed has a crucial influence. This can be measured using a tachogenerator. Determining the speed from other measured quantities, such as current and voltage, is also possible. One possible implementation is the use of a flux observer / model.

[0034] The current grid angle and the amplitude of the grid voltage are measured. These serve as starting values. Subsequently, using the grid frequency, the progression of the grid voltage over the predicted time steps is calculated.

[0035] With all parameters now known, the model can be initialized. The calculation of the state variables for each ignition possibility then continues until a current zero crossing occurs.

[0036] After prediction, it is clear which ignition possibilities exist in the current situation, i.e., which ignition options. These can then be checked and evaluated against a variety of criteria. The following decision criteria and corresponding combinations thereof are conceivable: 1. Maximum stator current, 2. Maximum torque peak / maximum torque amplitude, 3. Minimum conduction time for thyristors (to ensure defined ignitions and prevent misfires), 4. Minimum average torque, 5. Maximum rotor flux, 6. Minimum rotor flux growth due to ignition, 7. Thyristor losses, 8. Motor losses.

[0037] This list of possible decision criteria is not exhaustive. Further criteria may be used for decision-making, or certain of the above-mentioned possible decision criteria may be disregarded or weighted differently depending on the application under consideration.

[0038] These decision criteria can be checked either after or during the prediction process, especially for simple thresholds like maximum current or maximum torque. Ultimately, all possible ignition scenarios are evaluated, and a decision is made as to whether any of them are advantageous. If none of the possibilities appear conducive to achieving the desired goal, such as accelerating the engine, no action is taken, and the next iteration is awaited.

[0039] In principle, the decision criterion can directly influence the system's behavior. Operating modes are also conceivable that are not possible with current standard methods. By skillfully manipulating the maximum torque, a type of control system could be developed that enables ramped acceleration.

[0040] The entire process of prediction and decision-making can be tracked in a fixed or variable time frame, thus further accelerating the engine.

[0041] Advantageous embodiments and further developments of the invention are specified in the dependent claims.

[0042] According to a preferred embodiment of the invention, ignition possibilities are determined for each subsequent time step. The calculation times can be situated within a fixed time grid; in this case, the calculation times follow one another in defined, preferably uniform, time steps.

[0043] According to a preferred embodiment of the invention, the electrical and / or mechanical motor behavior is modeled using a transient motor model.

[0044] According to a preferred embodiment of the invention, the decision as to whether a calculated ignition option should be selected is based on one or more of the following decision criteria: Maximum stator current, Maximum torque, Minimum conduction time for thyristors, Minimum average torque, Maximum rotor flux, Minimum rotor flux growth due to ignition, Thyristor losses, Motor losses.

[0045] According to a preferred embodiment of the invention, no ignition takes place at the immediately next calculation time, and the prediction for a calculation time following the immediately next calculation time is awaited if the decision has been made that no ignition possibility is meaningful.

[0046] According to a preferred embodiment of the invention, only ignition possibilities are identified that result in negative torques and thus actively brake the asynchronous motor.

[0047] According to a preferred embodiment of the invention, a current rotor angle, a current rotor speed, a current network phase angle, and the current stator phase currents are repeatedly determined. Using the current rotor angle, the current rotor speed, the current network phase angle, and the current stator phase currents, a torque curve of a torque acting on the rotor is predicted for first ignition possibilities, in which two phases are switched to conducting by triggering thyristors, and for a second ignition possibility, in which three phases are switched to conducting by triggering thyristors. Using the predicted torque curves, a decision is made for each thyristor pair as to whether it is triggered.

[0048] According to a preferred embodiment of the invention, one or more iterations are omitted. The advantage of this is that the computational processor is relieved of some of its workload if it is currently engaged in other tasks, e.g., communication.

[0049] According to a preferred embodiment of the invention, one or more iterations are inserted, i.e., the iteration density is increased, e.g., under lower processor load or with few ignition opportunities. The advantage of this is that the processor's capacity is fully utilized to optimize the startup process.

[0050] According to a preferred embodiment of the invention, the step of predicting the electrical and / or mechanical behavior of the motor is preceded by a step of generating magnetic flux in the rotor by triggering the thyristors. In a permanent magnet motor, the rotor magnetic flux, also simply referred to as rotor flux, which is essential for generating the motor's torque, is generally generated by permanent magnets. This rotor flux is fixed in its amplitude and also fixed with respect to its mechanical position on the rotor. In contrast, in an asynchronous motor, no flux is initially present in the motor when it is stationary. The rotor is initially just a structure made of electrical steel, aluminum, and optionally copper.Therefore, magnetic flux, also simply called flux, must first be introduced into the machine, as this has a significant influence on the torque that can be generated by the asynchronous motor. This is achieved by introducing flux into the motor using one or, preferably, several current pulses, in order to subsequently accelerate the motor. Thus, with an asynchronous motor, it is necessary to introduce the flux into the motor itself. At the same time, the decision as to whether an ignition is worthwhile is no longer solely determined by whether torque or current limits are exceeded, but also by how the flux in the motor is affected by the ignition. If an ignition produces a particularly high torque but causes the flux in the motor to drop, a short-term gain may be achieved, but at the cost of only a reduced or relatively small torque being achievable in the next ignition, since the rotor flux is relatively low.The decision must therefore take into account whether the ignition timing chosen based on current and torque curves is also appropriate with regard to the motor's magnetization. This can be considered in the decision criterion, for example, by setting limits for "Maximum rotor flux", "Minimum rotor flux", or "Minimum increase in rotor flux during an ignition".

[0051] It is possible to generate a magnetized rotor, i.e., a magnetic flux in the motor's rotor, through two-phase ignitions at a consistent grid phase angle, meaning a consistent grid angle. These two-phase ignitions result in a pulsating current-space vector that imprints a magnetic flux in a precisely defined direction within the rotor. This rotor flux generation phase precedes the actual starting process to enable starting with a magnetized motor.

[0052] Another difference between a permanent magnet motor and an asynchronous motor is that, unlike the permanent magnet motor, the asynchronous motor requires consideration not only of the position but also of the variable amplitude of the magnetic flux. This is necessary because the amplitude also depends on the previous current application. Therefore, a method is required that determines not only the flux angle but the entire flux vector. Numerous methods for estimating the rotor flux in an asynchronous motor have become established in the prior art.

[0053] The fact that flux conservation must be considered in the decision criterion can be advantageously used as an additional degree of freedom. For example, it is possible to rotate the flux from its current position relative to the rotor to another position if the externally applied current causes this. This means that the rotor flux is no longer fixed at or in the rotor position, but can also rotate a certain distance; this is, incidentally, the reason why the motor always runs just below synchronous speed, i.e., asynchronously, when connected to the grid.

[0054] Furthermore, it is possible to increase or decrease the rotor flux through targeted ignitions, thus optimizing the start-up process. For example, by keeping the flux high through targeted manipulation of the decision criterion, very high starting torques, especially in the lower speed range, can be achieved. Conversely, it is possible to keep the flux at a relatively low value to minimize the corresponding rotor losses. This can be advantageous, for example, in applications such as fans. Here, relatively low torque requirements exist in the lower speed range (quadratic load characteristic); a high flux for generating high torques would not be necessary. Since fewer losses result from the targeted reduction of the rotor flux, more starts per unit of time are possible.

[0055] The problem is also solved by a soft starter according to the invention. The soft starter comprises an ignition signal unit for generating ignition signals and the thyristors. The ignition signal unit can be triggered to generate ignition signals for one or more of the thyristors. The soft starter also comprises means suitable for carrying out the steps of the described method. These means can be, for example, a control unit comprising a processing unit and a storage unit. A computer program can be loaded into the storage unit and permanently stored there, from where it can be loaded into the processing unit for execution.

[0056] The outlined problem is also solved by a computer program product according to the invention and a computer-readable medium on which the computer program product is stored. The computer program product is designed to be executable in at least one processor. The computer program product can be stored as software, e.g., as an app downloadable from the internet, or as firmware in memory and designed to be executable by a processor or an arithmetic logic unit (ALU). Alternatively or additionally, the computer program product can also be designed, at least partially, as a hard-wired circuit, for example, as an ASIC (Application-Specific Integrated Circuit). The computer program product according to the invention comprises instructions that cause the soft starter according to the invention to execute the process steps of the described method.The computer program product is thus configured to carry out the method for operating an asynchronous motor using a soft starter. In particular, it is configured to perform the step of determining possible ignition possibilities of one or more thyristors of the soft starter at a future calculation time. Furthermore, it is configured to perform the step of predicting the electrical and / or mechanical motor behavior if ignition of one or more thyristors of the soft starter were to occur, for the determined ignition possibilities. It is also configured to perform the step of deciding, based on the predicted motor behavior, whether and which ignition possibility should be selected. In addition, the computer program product includes commands that cause the ignition signal unit of the soft starter according to the invention to send one or more ignition signals for one or more thyristors.Several thyristors are generated if a decision has been made to enable ignition. According to the invention, the computer program product is designed to implement and execute at least one embodiment of the outlined method. The computer program product can integrate all sub-functions of the method, i.e., be monolithic. Alternatively, the computer program product can be segmented, distributing sub-functions across segments that are executed on separate hardware. Thus, the computer program product can be partially executable in a control unit of the soft starter and partially in an external control unit. Furthermore, part of the method can be performed in a soft starter device and another part in a control unit superior to the soft starter device, such as a PLC, a handheld parameterization device, or a computer cloud.

[0057] The properties, features, and advantages of this invention described above, as well as the manner in which they are achieved, will become clearer and more readily understandable in connection with the following description of the exemplary embodiments, which are explained in more detail in conjunction with the drawings. These show, in schematic representation: FIG 4 a soft starter according to an embodiment of the invention; FIG 6a a start-up with the conventional starting algorithm with a constant load torque ML of 30 Nm; FIG 6 a start-up with an embodiment of the predictive starting algorithm according to the invention with a constant load torque ML of 30 Nm; FIG 7a a start-up with the conventional starting algorithm with a constant load torque ML of 70 Nm; FIG 7a a start-up with an embodiment of the predictive starting algorithm according to the invention with a constant load torque ML of 70 Nm; and FIG 8 a flow diagram according to an embodiment of the method according to the invention.

[0058] FIG 4 Figure 1 shows the structure of a soft starter 1 according to an embodiment of the invention. The soft starter 1 can, for example, be configured in a topology such as that shown in Figure 1. FIG 1 As shown, the soft starter 1 is used. It comprises a control unit 41 with a processing unit 46, also referred to as a processor, and a memory unit 43. The processing unit 46 executes a computer program stored in the memory unit 43, which contains an algorithm for carrying out the procedure. During execution of the algorithm, ignition possibilities for at least one subsequent time step are determined, the corresponding electrical and / or mechanical motor behavior is predicted using a motor model, and finally, based on the predicted motor behavior, a decision is made as to whether and which ignition possibility should be selected. To initialize the algorithm, state variables of the system are measured or calculated. The processing unit 46 receives a series of measured values ​​44, e.g., the motor current I1,RMS, the motor voltage U1,RMS, and the rotor speed n, as input values.Grid voltages uA, uB, uC are measured to calculate the grid angle φgrid and the grid voltage amplitude Ugrid. The motor currents ia, ib, ic and motor voltages ua, ub, uc are also used for a model-based calculation of the rotor flux. Together with the measured mechanical rotor speed n and the grid frequency fgrid, all system variables are defined. After prediction, the possible ignition points at the considered future time are determined. These ignition points can then be checked and evaluated against a variety of decision criteria, such as a maximum torque or a maximum stator current.If a decision has been made for a specific ignition option, the control unit 41 generates one or more trigger signals 47 to an ignition block 42 of the soft starter, which includes thyristors and an ignition signal unit 48 for generating ignition signals for the thyristors. The trigger signals 47 cause the ignition signal unit 48 to generate ignition signals for one or more thyristors, thus establishing the pre-calculated engine behavior.

[0059] In the following, a standard model for induction motors is used as an embodiment of the method according to the invention for prediction. It should be noted that the motor model presented here is by no means the only one required to predict the motor's behavior. Simplified models that, for example, neglect factors such as stray fluxes, consider current displacement effects in the rotor, or use other quantities as state variables are also applicable. For the calculation of two-phase ignitions, the model states must be appropriately inverted, or the missing third-line voltage must be calculated separately.

[0060] The subscript 1 indicates that a value is a stator parameter, the subscript 2 that a value is a rotor parameter. The superscript K indicates that a value has been rotated in the reference frame by the rotation angle φ K. I → 1 K = I → 1 S ⋅ e − jφ K

[0061] The motor model includes the stator resistance R₁, the stator leakage inductances L₁σ, the mutual inductance Lₘ, the rotor resistance R₂, and the rotor leakage inductances L₂σ. For simplicity, the rotor parameters are referenced to the stator side. The electrical rotor speed is defined as Ω₁L, and the rotational speed of the reference frame as Ωₖ. The inductances are defined as: L 1 = L 1 σ + L h L 2 = L 2 σ + L h

[0062] Equations (6) and (7) show a model of a general induction motor based on the rotor-magnetic flux linkage. Ψ → 2 K and the stator magnetic flux linkage Ψ → 1 K as state variables: U → 1 K = R 1 ⋅ I → 1 K + d Ψ → 1 K dt + j ⋅ Ω K ⋅ Ψ → 1 K U → 2 K = R 2 ⋅ I → 2 K + d Ψ → 2 K dt + j ⋅ Ω K − Ω L ⋅ Ψ → 2 K

[0063] The magnetic flux linkage can be calculated based on the stator and rotor currents: Ψ → 1 K = L 1 ⋅ I → 1 K + L h ⋅ I → 2 K Ψ → 2 K = L h ⋅ I → 1 K + L 2 ⋅ I → 2 K

[0064] Engine torque is defined as: M M = 3 2 ⋅ p ⋅ Ψ → 1 K × I → 1 K

[0065] Assuming a squirrel-cage induction motor, a short circuit can be assumed on the rotor side: U → 2 K = 0

[0066] Furthermore, continuously rotating the reference frame for predictive calculations, as is often done in engine control, offers no advantage. Therefore, the rotational speed of the reference system is set to zero. dφ K dt = Ω K = 0

[0067] For a compact description of the model, it is also useful to define the dispersion factor σ: σ = 1 − L h 2 L 1 ⋅ L 2

[0068] From equation (9) it follows: I → 2 K = − L h L 2 ⋅ I → 1 K + 1 L 2 ⋅ Ψ → 2 K

[0069] In equations (6) and (7) the rotor-magnetic flux linkage Ψ → 2 K and the stator-magnetic flux linkage Ψ → 1 K defined as state variables. In the soft starter according to FIG 1 Only the stator currents are measured. These currents are a system property necessary for initializing the prediction model. To reduce the computational effort required to calculate the magnetic flux based on the measured currents, it is advantageous to have the current-space vector as a stator state variable. Therefore, equations (6) and (9) are used to calculate the stator current. I → 1 K to obtain as a new system-dynamic state variable: d I → 1 K dt = 1 σ ⋅ L 1 ⋅ U → 1 K − R 1 σ ⋅ L 1 ⋅ I → 1 K + R 2 ⋅ L h σ ⋅ L 1 ⋅ L 2 ⋅ I → 2 K − j ⋅ Ω L σ ⋅ L 1 ⋅ Ψ → 2 K d Ψ → 2 K dt = − R 2 ⋅ I → 2 K + j ⋅ Ω L ⋅ Ψ → 2 K

[0070] To reduce the computational effort, equation (14) is integrated into equations (15) and (16): d I → 1 K dt = 1 σ ⋅ L 1 ⋅ U → 1 K − R 1 ⋅ L 2 2 − R 2 ⋅ L h 2 σ ⋅ L 1 ⋅ L 2 2 ⋅ I → 1 K + R 2 ⋅ L h σ ⋅ L 1 ⋅ L 2 2 ⋅ Ψ → 2 K − j ⋅ Ω L σ ⋅ L 1 ⋅ Ψ → 2 K d Ψ → 2 K dt = R 2 ⋅ L h L 2 ⋅ I → 1 K − R 2 L 2 ⋅ Ψ → 2 K + j ⋅ Ω L ⋅ Ψ → 2 K

[0071] The torque equation (10) is transformed into the new system state variables using equations (8) and (9): M M = 3 2 ⋅ p ⋅ L h L 2 ⋅ Ψ → 2 K × I → 1 K

[0072] Equations (17), (18) and (19) represent the entire system model. For a 3-phase ignition case, this model can be used to calculate the electrical and mechanical motor behavior by applying the mains voltage. U → grid S as stator voltage U → 1 K is taken.

[0073] Calculating a two-phase trigger based on this model is also possible, but requires additional effort. For example, if the thyristors are triggered in phases a and b, the voltage across the thyristor in phase c must be calculated in parallel. This calculation must satisfy the condition ic = 0, because the thyristor in phase c is still blocking the current. Therefore, the calculation becomes even more time-consuming, as the voltage must be calculated in each prediction step.

[0074] To reduce this computational effort, the following approach can be used. When the thyristors are triggered in two of the three mains phases, they become conductive, whereas the thyristor in the third phase remains inactive. Because the sum of the currents in all three phases must always be zero, the current flowing towards the motor in one of the conducting phases must be identical to the current flowing away from the motor in the other conducting phase; therefore, the currents in the two conducting phases are exactly the same, only with opposite signs. FIG 5a und FIG 5b This concerns a two-phase ignition of phases a and b. In FIG 5a Within a fixed α-β frame of reference, it is obvious that this condition leads to a pulsating current vector that varies in length but not in direction. Therefore, only the absolute value of the current vector varies. I → 1 K . FIG 5b shows a graph with the absolute values ​​of the generated current vector.

[0075] If the direction is fixed and only the absolute value of the current state vector varies, it is no longer necessary to describe the current as a complex variable. By rotating all system state variables by an angle φP, which depends on the firing combination, the equations can be simplified. A prediction frame with the rotation angle φP is defined for this rotation: I → 1 P = I → 1 S ⋅ e − jφ P

[0076] The system equations (17), (18) and (19) are defined in the prediction framework as follows: d I → 1 P dt = 1 σ ⋅ L 1 ⋅ U → 1 K − R 1 ⋅ L 2 2 − R 2 ⋅ L h 2 σ ⋅ L 1 ⋅ L 2 2 ⋅ I → 1 P + R 2 ⋅ L h σ ⋅ L 1 ⋅ L 2 2 ⋅ Ψ → 2 P − j ⋅ Ω L σ ⋅ L 1 ⋅ Ψ → 2 P d Ψ → 2 P dt = R 2 ⋅ L h L 2 ⋅ I → 1 P − R 2 L 2 ⋅ Ψ → 2 P + j ⋅ Ω L ⋅ Ψ → 2 P M M = 3 2 ⋅ p ⋅ L h L 2 ⋅ Ψ → 2 P × I → 1 P

[0077] The rotation angle φP depends on the firing combination in the thyristors. For the three possible two-phase firing combinations: 1. Ignition of phases a and b 2. Ignition of phases b and c 3. Ignition of phases a and c are the rotation angles φ P within the prediction framework: φP,a−b=16π,φP,b−c=12π,φP,c−a=−56π

[0078] This method has several advantages compared to calculating the third voltage. Firstly, the third additional voltage of the no-current phase does not need to be calculated for each prediction step. Secondly, the motor model itself becomes simpler because equation (17) does not need to be solved twice, once for the real part and once for the imaginary part. Due to the rotation around the constant rotation angle φP, only the real part remains in equation (17) for calculating the two-phase ignition. If one of the conduction states of one of the thyristors is changed, the states of equations (17) and (18) and the mains input voltage vector U must be adjusted. grid rotated by the new angle φ P.

[0079] The actual rotor-magnetic flux linkage is used to calculate the electrical and mechanical behavior of the motor. Ψ → 2 K as a state variable of the motor, see equation (18). Because the measurement of the rotor-magnetic flux linkage Ψ → 2 K Since this is not possible with the standard hardware of a soft starter, a magnetic flux model is required to calculate the rotor-magnetic flux linkage. Therefore, equation (18) is used with the measured stator currents. I → 1 K The rotor magnetic flux linkage is implemented as an input value for tracking the rotor magnetic flux linkage. The rotor magnetic flux linkage value is also used in the next section to calculate a torque and a magnetic flux generation current (I1q and I1d). Together, they represent the stator current state vector, while φK is equal to the angle of Ψ → 2 P is: I → 1 K = I 1 q + jI 1 d

[0080] Using this approach, the torque equation (10) can be simplified: M M = 3 2 ⋅ p ⋅ Ψ 1 d ⋅ I 1 q

[0081] All calculations assume that the rotor speed Ω L is constant up to the prediction horizon. This assumption leads to acceptably small prediction errors because the inertia of the motor and load is, in most cases, large enough to keep the rotor speed stable up to the prediction horizon of less than 20 ms.

[0082] In the prediction step, all possible ignition sequences are calculated. These different ignition sequences result in different torque and current profiles over time. Based on these profiles and other decision criteria, a decision must now be made as to whether a particular ignition sequence is advantageous or not.

[0083] The primary reason for using a soft starter is to limit current and torque, preventing damage to the grid and application during motor start-up. Therefore, the decision criteria should include a maximum current amplitude ip,max and a maximum torque amplitude M p,max. Additionally, a minimum average torque M p,avgmin ensures that only firing possibilities that accelerate the rotor are accepted. Due to dead times in the firing hardware and inaccuracies in voltage zero-crossing detection, it is also useful to define a minimum conduction time tp,mc for the thyristors. Firing possibilities where conduction ends before this minimum conduction time tp,mc has elapsed are rejected.

[0084] Additionally, it is necessary to determine the amplitude of the rotor-magnetic flux linkage. Ψ → 2 K To maintain a usable level of magnetic flux, Ψp,free, and thus the rotor magnetic flux is generated from the stator currents. A minimum value is defined to ensure a specific level of magnetic flux in the rotor. Below this level, each ignition must increase the rotor magnetic flux by Ψp,Δmin. This increases the rotor magnetic flux until the minimum value Ψp,free is reached again. These rules allow the induction motor to be started up.

[0085] When the induction motor starts, the rotor is stationary and unmagnetized. To generate torque, a specific level of magnetic flux is required in the rotor. Therefore, at the start of the start-up, the thyristors are triggered one or more times using a two-phase firing sequence to generate a defined level of rotor magnetic flux (see diagram). FIG 5a und 5b . Measurements to demonstrate performance

[0086] The following section shows two measurements that illustrate the different behavior between the classical and predictive methods.

[0087] This measurement shows a start using a classic starting procedure. The current flows through the specified ignition angle in a recurring symmetrical pattern. The engine speed also increases continuously.

[0088] This measurement shows a start using a predictive method. Due to the specific decision criterion, the thyristors behave completely differently. First, the rotor is magnetized, characterized by uniform current pulses at the beginning. Then, after approximately 180 ms, the actual starting algorithm begins. Here, the current is unsteady and occurs almost in "packets." The acceleration, especially at low speeds, also resembles a staircase. Considering that conduction losses in thyristors only occur when a current is flowing, it becomes clear that the losses in the thyristors are significantly lower during this start than in the comparable classical starting process. The same applies, at least, to the stator winding of the motor.

[0089] FIG 6a and 6bshow a comparison of the behavior of a conventional and an algorithm according to the invention, based on a soft-start motor system according to FIG 1 The voltage and current measurements are taken using Hall sensors connected to an RCP system, which includes a CPU for executing the algorithm (CPU = Central Processing Unit; RCP = Rapid Controller Prototyping). The measured values ​​shown were acquired using a standard multi-channel power meter for measuring physical values. The rotor speed is measured with an inductive position sensor. Table 1 contains the motor data for the standard squirrel-cage induction motor used.

[0090] Table 1: Motor data of the standard model of squirrel-cage induction motor used nominal power, P N 15 kW nominal voltage, U N 230 V nominal current, U N 28.5 A nominal frequency, f N 50 Hz number of pole pairs, p 2 inertia motor, J M 0.0850 kgm 2< load inertia, J L 0.0955 kgm 2< stator DC resistance, R 1 0.150 Ohm

[0091] The induction motor is mechanically connected to a permanent magnet servo motor, which is powered by a high-performance servo inverter. This inverter allows for the simulation of mechanical loads with various characteristics. The moment of inertia from the servo load simulation is relatively small compared to typical industrial applications; this results in a relatively fast motor start-up and some oscillations, particularly noticeable during classical algorithm measurements. FIG 6a and 7a .

[0092] FIG 6a shows a measurement of the start-up of a soft-start motor system using a conventional algorithm and driving an application with a constant load torque ML of 30 Nm.

[0093] The voltage is set by a ramp generator, which starts with a relative starting voltage of 55% of the nominal motor voltage UN. The voltage then increases at a continuous rate of 25% per second. The inertial mass of the motor and load is accelerated by the difference between the motor torque MM and the load torque ML. This acceleration increases with the motor speed, i.e., the rotational speed n, and has a more or less constant form: the rotational speed n increases continuously. The current profile ia, ib, ic is a continuously repeating sequence that differs only in the firing angle and some oscillations. The current flows through the predetermined firing angle in a recurring symmetrical pattern. Additionally, the d-currents I1d and q-currents I1q are specified in the rotor-magnetic flux-oriented stator currents.These values ​​are calculated from the measured stator currents and the reconstructed flux values ​​of the rotor magnetic flux model. The current vector is divided into the field-generating current I1d and the torque-generating current I1q. Neglecting oscillations on the currents, the field-generating current is higher than the torque-generating current. Because the maximum current value per phase is more or less identical, the mean values ​​of the d and q currents also remain constant.

[0094] Equation (26) shows that the current motor torque MM depends on the q-current and the level of the rotor magnetic flux. Even if the q-current remains at the same level, more or less torque can be generated by influencing the rotor magnetic flux. This is the reason for the increasing acceleration at higher rotor speeds without a significantly higher q-current.

[0095] FIG 7a This shows another ramp-up using the conventional algorithm, with a higher constant load torque ML of 70 Nm. The other boundary conditions are the same as in the previous example. FIG 6a described.

[0096] FIG 6b Figure 1 shows a measurement of the start-up of a soft-start motor system, using a prediction algorithm according to the invention. The load and network conditions correspond to those of the conventional algorithm of [reference missing]. FIG 6a The load torque ML is a constant 30 Nm.

[0097] Before the prediction algorithm starts, a special section is added to the startup sequence: First, the rotor is magnetized, characterized by uniform current pulses at the beginning. This involves several two-phase ignitions to establish the rotor-magnetic flux linkage. Ψ → 2 K to generate. This leads to a current space vector whose direction is fixed, but whose amplitude pulsates, see FIG 5a und 5b , which determines the amplitude of the rotor-magnetic flux linkage Ψ → 2 K The current is continuously increased. Because only rotor magnetic flux and no motor torque MM is generated, a pure d-current is present in this section of the start-up sequence. During the time interval from t=0 s to t=0.18 s, it serves as a prerequisite for torque generation in the next part of the start-up sequence.

[0098] The algorithm according to the invention then begins to run: Here, the current is intermittent and occurs in bursts. Compared to the start-up with the conventional algorithm, there is no continuous ignition profile, but rather alternating periods with and without current. This also leads to a difference in the acceleration behavior. The acceleration is not as smooth as with a soft starter controlled by a conventional algorithm, but has a stepped form, especially at low speeds: There are periods during start-up in which the acceleration is relatively high (periods with current flow), and periods without acceleration (periods without current flow). Furthermore, the d- and q-currents exhibit a different shape. Here, too, there are periods with and without current.

[0099] Furthermore, the relationship between the levels of the d- and q-currents differs from that observed during conventional startup. The d-current is much lower compared to the conventional algorithm. It is also lower than during startup using the conventional algorithm. This indicates that the chained rotor magnetic flux is higher when using the prediction algorithm.

[0100] FIG 7b This shows another ramp-up using the prediction algorithm, with a higher constant load torque ML of 70 Nm. The main parts of the ramp-up sequence are similar to those of the sequence using the prediction algorithm with the lower load torque ML of 30 Nm. The increase in the absolute rotor-magnetic flux linkage occurs in the time interval from t=0 s to t=0.18 s. Ψ → 2 K . After that, the same pulsating behavior can be observed as in FIG 6b However, from t=0.88 s until synchronization at t=1.1 s, a behavior is observed that differs significantly from the behavior before. During this ramp-up period, the current profile changes, and the current profile resembles that of the conventional algorithm. This is only possible because the requirements of the decision criteria are also met by the behavior of the conventional algorithm. Before this ramp-up period, the firings using the conventional algorithm do not match the decision criteria. For example, the minimum magnetic flux criterion cannot be met with a single firing using the conventional algorithm. The prediction algorithm achieves the same result in a different way than the simpler firing-angle-based solution because the conventional solution fulfills the requirements of the decision criteria in the final ramp-up period.

[0101] Comparing the current curves of the conventional algorithm and the prediction algorithm according to the invention, it becomes clear that the prediction algorithm leads to a start-up with a shorter current conduction time. This also results in lower losses in the thyristors of the soft starter and the current lines of the induction motor. A simple model based on the forward voltage Uf and the thyristor dynamic resistance Ron is used to estimate the losses Pthy in the thyristors: P thy I thy = I thy ⋅ U f + R on ⋅ I thy 2

[0102] The parameters of the thyristor module used are summarized in Table 2. Table 2: Technical data of the thyristor module used forward voltage, U f 0.9 V thyristor dynamic resistance, R on 2 mOhm

[0103] For the sake of simplicity, only the resistive stator losses are compared, because they are relatively easy to calculate using the resistance R1 of the stator windings.

[0104] Tables 3 and 4 show the calculated losses, separated as thyristor losses and resistive stator losses, for a motor start-up with a constant load torque of 30 Nm and 70 Nm, for a start-up using a conventional algorithm and a start-up using the predictive algorithm. The values ​​were based on the data in the FIG 5a, 5b , 6a and 6b The measurements shown are calculated. The energy measurement period begins with the first ignition signal and ends when a steady current is reached. Table 3: Calculated losses for the measurement with a constant load torque of 30 Nm Classic Start-up Predictive Start-up Thyristor losses 213 J 108 J Resistive stator losses 3557 J 1741 J Table 4: Calculated losses for the measurement with a constant load torque of 70 Nm Classic Start-up Predictive Start-up Thyristor losses 414 J 209 J Resistive stator losses 7614 J 3585 J

[0105] Tables 3 and 4 show that the thyristor losses can be reduced by almost 50 percent with the predictive algorithm under both load conditions compared to the conventional algorithm. The resistive stator losses were also halved.

[0106] Both cases demonstrate that the invention makes it possible to reduce losses in the soft starter and the induction motor during the starting process. This makes it possible to perform more starts per unit of time, reduce cooling requirements, and thus build smaller soft starters and save energy when an application is started frequently.

[0107] Figur 8 shows a flowchart of a procedure with process steps S1 to S4 for operating an ASM with a soft starter.

[0108] In a first process step S1, possible ignition possibilities of one or more thyristors of the soft starter are determined at a future calculation point in time. After the first process step S1, a second process step S2 is executed. In the second process step S2, the electrical and / or mechanical motor behavior, if ignition of one or more thyristors of the soft starter were to occur, is predicted for the determined ignition possibilities. After the second process step S2, a third process step S3 is executed.

[0109] In the third process step S3, a decision is made, using the predicted engine behavior, as to whether and which ignition option should be selected.

[0110] In the fourth process step S4, one or more firing signals are generated for one or more thyristors, if a decision has been made regarding a firing option. After the fourth process step S4, the process continues with the first process step S1.

Claims

1. Method for starting and stopping an asynchronous motor (4) by means of a soft starter (1), having the following steps: - ascertaining firing opportunities possible for one or more thyristors (2) of the soft starter (1) at a calculation time in the future; - forecasting for the ascertained firing opportunities the electrical and / or mechanical motor behaviour if a firing of one or more thyristors (2) of the soft starter (1) were performed; - taking the forecast motor behaviour as a basis for deciding whether and which firing opportunity is supposed to be chosen; and - generating one or more firing signals for one or more thyristors if the decision has been made to take a firing opportunity, characterized in that - a present rotor angle, a present rotor speed, a present grid phase angle and the present stator phase currents are ascertained repeatedly, - the present rotor angle, the present rotor speed, the present grid phase angle and the present stator phase currents are used to anticipate a torque response for a torque acting on the rotor for first firing opportunities, in which firing of thyristors turns on two phases, and for a second firing opportunity, in which firing of thyristors turns on three phases, and - the anticipated torque responses are used to decide whether each thyristor is fired.

2. Method according to Claim 1, wherein firing opportunities are ascertained for the next particular time step.

3. Method according to Claim 1 or 2, wherein the motor behaviour is modelled by means of a transient motor model.

4. Method according to one of the preceding claims, wherein the decision is made on the basis of one or more of the following decision criteria: maximum stator current, maximum torque, minimum on-time for thyristors, minimum average torque, maximum rotor flux, minimum rotor flux growth as a result of the firing, thyristor losses, motor losses.

5. Method according to one of the preceding claims, wherein firing does not take place at the immediate next calculation time and the forecast for a calculation time that follows the immediate next calculation time is awaited if it has been decided that a firing opportunity is not useful.

6. Method according to one of the preceding claims, wherein only firing opportunities that result in negative torques and therefore actively slow down the asynchronous motor (4) are ascertained.

7. Method according to one of the preceding claims, wherein the step of ascertaining firing opportunities is preceded by a step of magnetic flux generation in the rotor by way of firings of the thyristors.

8. Method according to Claim 7, wherein the magnetic flux in the rotor is generated by way of two-phase firings of the thyristors at always the same grid phase angle.

9. Soft starter (1) having a firing signal unit (48) and means (41) that are suitable for carrying out the steps of the method according to one of Claims 1 to 8.

10. Computer program product, comprising instructions that cause the soft starter (1) of Claim 9 to carry out the method steps according to one of Claims 1 to 8.

11. Computer-readable medium on which the computer program product according to Claim 10 is stored.