1. Introduction

Permanent Magnet Synchronous Motor (PMSM) is a synchronous motor whose rotor uses permanent magnets instead of windings.

PMSM uses permanent magnets to provide excitation, which makes the motor structure simpler, reduces processing and assembly costs, eliminates collector rings and brushes, and greatly improves the reliability of motor operation. Because no excitation current is required and there is no excitation loss, the efficiency and power density of the motor are improved.

At present, domestic electric vehicles are basically powered by permanent magnet synchronous motors. The use of SaberRD software can perfectly realize the simulation of electric vehicle power systems. This paper abstracts the electric vehicle power system from four levels.

The simulation objectives include global efficiency, thermal analysis of long-term driving, nanosecond inverter switching characteristics and loss simulation. Through simulation, motor and inverter control parameters can be optimized, power quality (THD and losses) can be verified, and faults can be simulated.

The four levels of abstraction mainly include:

All device characteristics mentioned in this article are based on published data (see Table 1 for details).

Table 1 Automobile power system parameters

The core of the power system design is the permanent magnet synchronous motor model, whose parameters include spatial harmonics, magnetic saturation and frequency-related losses. (See Figure 1 below)

Figure 1 Motor parameters

2. Level 1

Figure 2 Automotive motor drive system simulation circuit (Level 1)

The circuit simulation speed of layer 1 abstraction is the fastest and is used to simulate long-running projects. In Figure 3 below, the actual running time of 7 hours (until the battery is exhausted) can be completed in 60 seconds of simulation time. The motor and inverter use dq models to achieve the fastest simulation speed.

The dq model (frequency domain based) parameter acquisition is based on phase and amplitude data obtained by analyzing the high frequency switching characteristics of voltage and the sinusoidal nature of current using averaging techniques. This level of abstraction is very suitable for thermal analysis of long-term simulation.

The factors that affect the driving range of the car after multiple charge and discharge cycles are the losses of the inverter and the motor, which can be accurately obtained through a lookup table. These table information can be obtained from the IGBT model or the FEA test results of the motor.

The losses of the motor and inverter are frequency-dependent (inverter switching frequency and motor rotation frequency). To ensure sufficient electrothermal coupling, the inverter losses are also temperature-dependent. No temperature-dependent parameters are added to the motor model here.

The heat flow generated by the inverter and the motor can be connected to a simple thermal network at 20°C. Figure 3 shows the actual speed, temperature and battery voltage waveforms in the cycle continuous driving mode. The NEDC simulation can finally run a distance of 245km, while the manufacturer reports a distance of 200km. The reason for this phenomenon is that the simulation model is 100% ideal and there is no loss. (The full name of EDC is: New European Driving Cycle, which means "New European Driving Cycle" in Chinese. The mileage standard of the Ministry of Industry and Information Technology used in my country is the European standard. The NEDC mileage test mainly simulates the environment of urban and suburban areas, accounting for 4:1 respectively. Because there are many factors affecting the actual road conditions and environment, the NEDC test is basically a bench test.)

By setting up transient simulation, we can obtain the transient efficiency of the motor and inverter and plot the waveforms. Signal name plotting the signals instant_efficiency in vsi_dqx.vsi and jmag_pmsm_dqx pmsm:fea_pmsm_dqx.pmsm. The average efficiency of the inverter is 79% and the efficiency of the motor is 85%.

Figure 3 Simulation results

3. Level 2

Figure 4: Automobile motor drive system simulation circuit (Level 2)

Level 2 circuit, which changes the DC-based dq axis of Level 1 to the abc axis, uses a non-switching 3-phase inverter to generate synchronous sinusoidal voltage. The simulation time of Level 2 circuit will be slower. Because the periodic sinusoidal signal is not abstracted. However, the voltage is discontinuous, which is faster than the pulse width modulated Level 3 circuit. Level 2 circuit is a compromise between speed and accuracy, suitable for studying motor drive circuits. In particular, the torque pulsation caused by the motor space harmonics can be observed.

Figure 5 shows that the car accelerates from 0 to 60nph in about 10 seconds on a flat road. Setting the values ​​of the terrain attributes of the load_veh_dyn symbol to be equal can achieve the effect of running on a flat road.

Figure 5 Simulation results of a car on flat terrain

It can be seen from the above figure that the frequency of the phase current is proportional to the speed of the car, and the id value of the current vector is negative, which conforms to the MTPA formula.

Figure 6 shows that the back EMF generated by the motor is limited by the battery voltage at 5s. After that, the motor is in the weak magnetic control mode, which can further increase the speed, but the error between the theoretical value and the actual value will increase. The simulation results also show that the motor current is limited by the MTPA algorithm after reaching the set maximum of 300A.

Figure 6 Weak field control mode

Figures 7 and 8 show the simulation of vehicle dynamics in the ramp segment. In the downhill segment, the vehicle acceleration exceeds the set value, and the motor generates reverse torque, temporarily acting as a generator. At this point, energy flows back to the battery through dynamic, magnetic, electrical and chemical reactions, and the change in reverse current can be seen when the torque passes through zero.

Figure 7 Speed ​​response on slope terrain

Figure 8 Torque and current response on slope terrain

To prove this, change the load_veh_dyn symbol terrain attribute profile: [(0, 0), (1, 0), (10, -5)], simulate again, and observe the waveform.

It is worth pointing out that in order to obtain the parameters for the dc operating point, the initial position of the car must be slope-free. The motor has no speed when it starts, and no torque is required. Because the balance of forces acting on the car can only be calculated when there is no gravity, if you want to modify the slope parameters, make sure it is flat at x=0.

4. Level 3

This layer of circuit uses a PWM inverter model, the simulation speed will be slower, and the control algorithm is implemented using the MAST language.

Figure 9 Automobile motor drive system simulation circuit (Level 3)

Level 3 abstracts the PWM inverter operation and uses ideal switch models and diode models. The simulation speed is slower than Level 2. However, it is an order of magnitude faster than Level 4, which uses actual semiconductor models.

The PMSM template uses the MAST language template, using sampled signals instead of Level 2 continuous signals. Unlike the Level 2 output using duty cycle cycles, this method causes the inverter switches to gradually disappear, making it closer to a real MCU. However, most of the Level 3 operations are the same as Level 2, including Park transformation and Park inverse transformation, MTPA, field weakening control, PI integration and duty cycle calculation.

Instead of triggering the switching devices through asynchronous interrupts, the control system controls 12 switching devices at the beginning of each switching cycle (for 6 switches, each switch is turned on and off once). The switching sequence is then queued in the simulation event queue. In the communication scheme, when the system is simulating an external FPGA, virtualizer, etc., controlling the fixed and relatively long time interval (200us) can significantly improve the system simulation performance, which allows the switching events to be temporarily stopped without major losses.

If it is not used for collaborative simulation, the sampling frequency of the system can be set higher than the switching frequency. Increasing the sampling rate can use filters such as Kalman to reduce noise on the data. The switching frequency and sampling frequency switching time are mainly based on the duty cycle, which is the ratio of voltage synthesis based on the DC bus voltage.

The dead time specified by the user is also included in the duty cycle calculation. The dead time is a small interval between the closing and opening of two complementary IGBTs when they are turned on simultaneously to avoid short circuit of the inverter. The dead time is generally between 1 and 5us.

Figure 10 Corrected and uncorrected dead zone distortion

Figure 10 shows that without dead time compensation, the current waveform changes caused by dead time are very obvious when one phase current crosses zero (which happens six times per synchronous cycle). The waveform abnormality will lead to increased THD levels and motor core losses. At high-frequency PWM, since the dead time is fixed (depending on the internal switching speed of the IGBT), the duty cycle of the dead time in the switching cycle will increase, and its effect will be worse. At low-speed synchronous switching, the impact of dead time is also obvious.

The distortion of the dead time can be corrected and adjusted at the moment of current polarity switching. When the same pin of the IGBT is turned off, the related inductive current flows through either the upper freewheeling diode or the lower freewheeling diode. Because the current polarity of the pin is known in advance, it is possible to know in advance whether the phase voltage is zero voltage or bus voltage, which helps to increase or decrease the duty cycle in advance.