Insights into Motor Drive Design: Analyzing Trends and Long-Term Costs with PSIM and HyperStudy

The HyperStudy Advantage

Motor drives may seem like a complex puzzle, where the pieces require both a strong theoretical foundation and hands-on experience to fit together. Coming up with the ideal recipe for the motor's optimal performance is no easy feat, and it often requires that multiple parameters are considered at the same time. 

PSIM & HyperStudy can offer you a cost-effective express lane to insights derived from a multitude of simulations. For instance, you can tackle the ever-longing debate of MOSFETs versus IGBTs for your system by utilizing simulation data. You can visually see how the choice of switch, switching frequency and system efficiency are connected:

Navigating the PSIM - HyperStudy Workflow

You can find a step-by-step tutorial for connecting HyperStudy and PSIM here: Linking PSIM with HyperStudy

The goal is straightforward: We leverage the PSIM engine for multiple parallel simulations of our motor drive, gathering data. Our focus isn't on absolute simulation fidelity; instead, we seek an overview of the system's dynamics by covering various interactions.

"Instead of concentrating all your efforts on a single high-fidelity simulation, you can now explore trends to drive actionable decisions"

This doesn't mean that we will be using ideal models for the key components of the system. Instead, by utilizing PSIM's thermal module, we can have accurate inverter loss results while maintaining a decent simulation speed. The same applies for the Flux and FluxMotor dedicated motor models inside PSIM. These models utilize a simplified model derived from Flux or FluxMotor, thus preserving the FEA tool accuracy. They also enable simulation of custom motor designs. More details about the Flux/FluxMotor models can be found here: Analysis and design of complex motor drive systems

For this workflow, let's consider a FluxMotor PMSM Drive with speed control and a constant torque load:

_____________________________________________________________________________________________________________________

A "reconfigurable" inverter model can be used in order for the PSIM parameter file to be able to access:

1. If an IGBT or a MOSFET/ SiC/ GaN device will be used as the switch model

2. The actual part name of the model used

** To use a specific switch thermal model, it must first be created in the PSIM thermal models library. Users can build their library of devices by including their models of interest. Follow these links for tutorial videos on this topic : Introduction to the Thermal Module  and PSIM thermal module improvements

_____________________________________________________________________________________________________________________

After completing the power stage and control stage design in PSIM, the HyperStudy process begins. HyperStudy can directly access PSIM's parameter file to define new parameters for each simulation in the automated Design of Experiments (DOE). Additionally, HyperStudy can retrieve PSIM's simulation results, capturing every signal that is connected to a probe or is current/voltage flagged:

_____________________________________________________________________________________________________________________

Key performance indicators (KPIs) to be considered in this specific workflow are:

1. Total System Losses [Efficiency]

2. Total Harmonic Distortion

3. Long term investment cost: Initial + Operational

The aim is to address questions such as:

"How do parameters like switching frequency and modulation technique influence the KPIs?"

"Is the choice of switch more critical than selecting the switching frequency for this particular system?"

We will answer those questions in a graphical manner with the help of HyperStudy:

Total Losses | Efficiency

Within HyperStudy, an array of graphical tools facilitates result post-processing. Some examples are: Scatter plots, Distribution plots, Pareto plots, Bubble plots, Interactions and of course Linear Effects plots.

Linear Effects Plots serve as a lens into system trends by concentrating results and comparing an averaged trend across a series of experiments. These plots segment results based on the X-axis input parameter and its impact on the Y-axis output KPI. The collective average points for each X-axis value from all experiments are interconnected, forming a line. This line displays the trend or rather the linear relation between a specific input parameter (like the switching frequency) and its influence on a KPI (like the total losses).

If the X-axis values are discrete, for example input switching frequencies of 8kHz - 12kHz - 16kHz - 20kHz, then the results will also appear discretized.

_____________________________________________________________________________________________________________________

Bubble plots allow you to inspect up to 4 different metrics at the same time:

_____________________________________________________________________________________________________________________

Crossings of interaction plots are also a very interesting assessment tool as they portray that the system behavior can be optimized differently between operation points:

It is important to highlight that this is not true for all motor drives, but rather applies to the characteristics of our own specific system. Each design requires its own unique investigation.

_____________________________________________________________________________________________________________________

Results can also be displayed side by side to examine the effects of all input parameters on a selected KPI, such as system efficiency:

1. Based on this graph, we conclude that the biggest effect on the efficiency comes with the switch selection

2. Switching frequency is the next big factor affecting the efficiency, but has almost half the effect of the SW selection and a comparable effect with that of the modulation technique. That being said, we can leverage DPWMmax or min with some higher fsw to save up on the losses. But the effect of the SW selection itself seems to be far more important than those choices.

3. Increasing the load torque has a higher toll into system efficiency than increasing the output speed. Hitting the same mechanical output power with larger speeds instead of more torque can lead to increased efficiency.

4. Dead time has minimal effect when it comes to system efficiency.

_____________________________________________________________________________________________________________________

THD

HyperStudy's outputs are not limited to what the PSIM simulation results have to offer. Custom Compose OML scripts and Python scripts can be created and easily added into HyperStudy to post process the PSIM results. Follow this link to inspect how easy it is to connect your scripts: Using Compose or Python Functions Within HyperStudy

For instance, a custom script was developed to compute the Total Harmonic Distortion (THD) of the motor's supply currents. Elevated THD levels can induce torque ripple, resulting in undesirable motor vibrations and noise.

_____________________________________________________________________________________________________________________

Long Term Investment Cost for the Inverter

Integrating post-process scripts into HyperStudy unlocks a whole new range of capabilities. Custom cost functions can factor in initial and operational costs (in this case tied to inverter losses) for a selected set of switch part numbers. These switches, ranging from various IGBTs to MOSFETs available in the market, can be compiled into a vendor data list. The script can then access this list to correlate pricing and other vendor information with PSIM's simulation results. 

Naturally, each device on the list should correspond to a PSIM thermal model (the PSIM model must be created in advance). And there you have it! HyperStudy links losses and costs to specific switch models.  

Moreover, users can input additional variables into the cost function:

1. Cost per kWh in their region

2. Annual working hours

3. The projected duration (in years) for long-term cost analysis

Let's examine the total 10-year costs of this system's inverter, considering both initial and operational expenses due to losses:

The cost function of this example is rather simple, just a basic payback financial model. We are not limited to this! We could have included discounted cashflows, inflation, and other more complex interactions for a more detailed financial model and to better determine the net present value (NPV) of the investment.

_____________________________________________________________________________________________________________________

Summing up the steps:

_____________________________________________________________________________________________________________________

This workflow showcases that a large enough dataset can act as your compass, helping you navigate through your system dynamics and costs. What is more, you have all the tools to create custom functions and evaluate your own KPIs.

As a quick reference, here are all the links shared within this blog:

Linking PSIM with HyperStudy

PSIM and HyperStudy for Multiphysics goal seeking - Optimize Performance and Cost 

Analysis and design of complex motor drive systems

Introduction to the Thermal Module

PSIM thermal module improvements

Using Compose or Python Functions Within HyperStudy