Power Electronics Development Center

cancel
Showing results for 
Search instead for 
Did you mean: 

A brief history of NI investments to provide a complete toolchain for digital energy control systems

NI CompactRIO and sbRIO General Purpose Inverter Controller

 

Although NI CompactRIO and LabVIEW FPGA have been used since 2003 in various power electronics and power system control applications, our first major hardware investment specifically targeted at providing the exact I/O mix and hardware specifications for power conversion equipment is an embedded system product called the NI General Purpose Inverter Controller (GPIC). It was developed in collaboration with NREL and several dozen of industrial power converter design teams around the world. The idea is to reduce the development time, cost and risk for grid tied power converters and motor drives greater than 50 kVA, while providing the benefits of full custom circuit design (via user programmable FPGA hardware and custom interface circuit board mechanicals) with the risk reduction of off-the-shelf hardware with long lifetime support, service (protection against end of life parts) and low cost upgradeability to the latest Moore's Law technology provided by NI.

 

Note that we continue to sell many CompactRIO based power converter control systems. The added per unit cost of the metal housing and mechanical shock & vibration ratings are minimal but the pre-ruggedized, EMC rated metal chassis based system reduces mechanical/electrical development complexity and system certication cost significantly. Also, the modularity of the C Series I/O provides added flexibility that enables a single chassis to become both an inverter control system and a synchrophasor measurement unit, for example.

 

As we studied the renewable energy industry and talked with many companies developing products for the smart grid, it became clear that the complexity of designing (and providing lifetime support for) electronic processing/control boards today is a major barrier and risk. With Moore's Law, the complexity of hardware continues to increase exponentially.

 

Hybrid FPGA/DSP Chipsets

 

A significant paradigm shift for power electronics is that the DSP cores have moved inside the FPGAs, which lowers the cost for a device with both capabilities by 20-70 fold. So that gives a major economic incentive to use the new hybrid FPGA/DSP devices rather than separate monolithic chipsets. However, the traditional hardware design tools for FPGAs (VHDL, Verilog) are extremely low level and tedious and lie outside the expertise of 99.9% of power electronics engineers. On the other hand, NI has been investing in high level graphical programming tools for FPGAs (including the new hybrid ones) for over 15 years, and our LabVIEW FPGA technology enables engineers with little or no programming experience and no knowledge of VHDL or Verilog to design their own FPGA hardware with signal processing and control performance that rivals traditional hardware description languages. These hybrid FPGA/DSP chipsets are also known as heterogenous systems on a chip.

 

Simulation Tools for Hybrid FPGA based Power Electronics Control Systems

 

With the availability of both single-board RIO and CompactRIO based deployment hardware containing hybrid FPGA/DSP chipsets, NI's main gap was a lack of tight simulation integration between the graphical FPGA programming tools and power electronics simulation tools. Also, our customers using competitive FPGA programming toolchains reported that their text based FPGA code (sometimes automatically generated) quickly became out of sync with their simulations (rendering the simulation results invalid), and about 80 percent of their FPGA software development cost was going towards I/O and communication bus interface code (rather than valuable control algorithms, protection logic, PWM, etc.). We knew we could resolve both of these issues with the right FPGA power electronics simulation tools and could enable our customers to go back and forth between simulation and compiled/deployed code running in FPGA hardware multiple times in a day rather than the whole process being a one way ticket (from simulation to VHDL) that takes days/weeks/months between design iterations.

 

Co-Variable Timestep Co-Simulation

 

About a decade ago, NI acquired Electronics Workbench, which develops the Multisim and Ultiboard circuit design tools. So the NI Multisim, NI LabVIEW Control Design & Simulation and LabVIEW FPGA teams began working together to turn Multisim into a first class simulator for power electronics circuitry, and developed a novel co-variable timestep simulation interface with LabVIEW, which effectively creates a continuous time co-simulation tool. This eliminates the need for the user to know what fixed timestep is appropriate for the entire simulation (an unreasonable requirement in our opinion) and enables the simulator to automatically slow down time as necessary to accurately simulate fast transient events (switching, short circuits, etc.) while automatically speeding up during more steady state conditions. This was based on feedback from the industry that most co-simulation tools don't accurately capture the transient, closed loop interaction between the analog world of power electronics and the digital world of PWM control systems, which particularly with modern FPGA-based control systems, can happen on nanosecond time scales. For industry power conversion design teams, the short circuit and fault events are the most important test cases to design the FPGA control system to handle properly.

 

We have also added thousands of power electronics parts from Infineon and other vendors to Multisim. However, for the high power IGBT modules, the vendors could not provide detailed simulation models for us. So, we solved the problem a different way by creating a highly requested generic Transistor_Diode model that can be populated with datasheet information to enable accurate transient thermal (junction-to-case-to-heatsink). Unlike the generic sizing and thermal simulation tools provided by the vendors, this enables development teams to conduct switching/conduction loss (energy efficiency) and transient thermal simulations (for sizing, cooling design and selecting protection limits) based on their exact FPGA pulse-width-modulation code and closed loop control algorithms, rather than a generic sine-triangle PWM scheme.

 

Enabling Control System Validation & Verification with Real-Time HIL Simulation

 

A glaring technology gap in the commercially available design tools for power electronics control is the lack of real-time HIL simulators that are sufficiently fast to enable validation and verification testing of power electronics and microgrid control systems. The automotive industry would never think of releasing an engine control unit (ECU) without the complete test coverage enabled by HIL testing, yet most commercial power electronics control units are released to customers with only the minimum level of validation enabled by physical testing. So for the last five years we've been working with our R&D teams to try to develop a toolchain for real-time HIL simulation of power electronics and fast transient power systems like microgrids and distribution grids.

 

Some of our inverter customers were actually building analog op-amp computers to enable real-time HIL validation of their inverters. The  basic rule of thumb is that the simulator needs to be at least 100  times faster than the PWM carrier frequency to achieve 2 percent  accuracy in capturing the switching events and simulating transients.The only digital hardware I'm aware of that can meet the fast input-to-simulator-to-output requirements for real-time power electronics with switching frequencies above 2 kHz is a hybrid FPGA with DSP cores woven into the programmable logic fabric.

 

Floating Point Math on Hybrid FPGA Hardware

 

Until recently, power electronics HIL simulators (and control systems) had to be programmed using fixed point math, which means that it could take 6 months of expert human labor to convert a simulation model into FPGA hardware. Worse, the end result is extremely brittle numerically. For example, changing a capacitor value from 1 microfarad to 1 millifarad or changing the input voltage from 240 VAC to 480 VAC may necessitate redoing (and retesting) all of the fixed point math selections.  So we came to the conclusion that it would be virtually impossible for NI to make a universal power electronics HIL simulator based on fixed point math.

 

Fortunately, in recent years floating point math has become feasible in hybrid FPGAs, so for the first time we can create universal simulators with the required MHz speed capabilities. We started with floating point state-space and transfer function solvers, and most recently have created a discrete time SPICE-like  circuit simulator. This is one of the most exciting projects I've worked on in my career at NI. We have a working prototype in which we can automatically export the simulation models from Multisim and fully automate the matrix inversions and translations necessary to load it into the floating-point, FPGA-based circuit simulator. I can change the schematic wiring or the value of a component, restart the simulation seconds later and it's running as fast as a real circuit. Right now it's just an alpha prototype but I'm hoping there will be sufficient interest from industry for us to convince other NI leadership to staff up and turn it into a full featured product. (Please email me if you'd like to request this, or visit with me at the APEC conference next week where we will be showing demonstrations- see screenshot below running on NI FlexRIO.) 

 

Multisim2FPGA.png

 

Electric Motor & Drive HIL Simulation with FEA Accuracy

 

We recently productized and released a real-time motor + inverter drive simulation toolkit that supports SRM and PMSM motors simulation at MHz speed. This includes medium fidelity models (sinusoidal flux, balanced 3-phase) and high fidelity models based on JMAG FEA, with a novel FPGA based solver that enables us to replicate the fidelity of FEA in the real-time simulation. (It's much more sophisticated than simple look up table methods.) Based on the beta program response, there seems to be a tremendous unfilled need for these kind of control system HIL validation tools. Subaru used the toolkit to achieve complete test coverage on their new XV Crosstrek Hybrid, including test conditions that would be impossible to reliably produce on a test track or dynamometer (case study).

 

Economic Impact

 

We are happy with the progress towards these goals that has been made in recent years, and the impact it is having on our customer's business such as at Dynapower Corporation. Based on independent survey results, taking an NI LabVIEW RIO approach to embedded systems design enables an average cost savings of $950,000 USD in engineering development cost per embedded systems design (114 person-months). A qaulitative survey of developers using the new Desktop Execution Node for LabVIEW FPGA co-simulation suggests it enables about a 10 fold productivity improvement in the model based design process.

 

From a human perspective, the NI LabVIEW FPGA graphical programming tools and patented co-simulation tools enable power system and power electronics domain experts, rather than low-level hardware design engineers, to design advanced algorithms for heterogeneous FPGA hardware and deploy them to pre-validated embedded targets like the NI GPIC and CompactRIO, which are designed for commercial deployment using high level LabVIEW FPGA graphical programming tools.

 

Thinking about Future Investments

 

Despite some progress, significant continued investments and deep collaboration with researchers in industry, government and academia will be required to complete the Design V Toolchain for networked distributed power grid control systems.

 

National Instruments is engaged in sustained R&D outlay to fulfill our vision of providing a comprehensive Design V Toolchain, present status shown in the figure below, for next generation digital energy control systems. Our goal is to significantly improve the model based design and simulation, commercial deployment, product validation and verification, production testing, real-time networking, and remote asset management of embedded digital energy control systems such as power conversion control equipment, grid tied inverters, variable frequency drives, and smart relay protection devices.

 

Your requests and feedback are vital to defining the right R&D and product development roadmap. How are we doing? How is it impacting your business or research? What are the gaps or weaknesses you perceive in the NI toolchain? In what areas should NI be investing to provide exactly the right embedded systems hardware and software tools to meet your current and future needs?

 

DesignV.jpg 

 

 

 

 

Message 1 of 4
(9,085 Views)

This is a great article. Today, the electric industry need more flexible and more affordable tools to simulate different power system phenomena as we integrate more renewable generation into the grid. It seems to be a little bit overwhelming to use simulation system and control hardware from the same vendor at the beginning. But if the development productivity is boosted significantly, the developers will embrace the concept and the process. I have enjoyed using NI platform to develop controllers because I can easily test the function at any stage of the development. Have you ever looking into to include the traditional EMTP type of simulation for traditional AC systems? Eventually, the converters have to be integrated into the AC network, at least for the foreseeable future. It is important to have a hardware-in-the-loop system that can simulate both AC and DC networks.

0 Kudos
Message 2 of 4
(5,996 Views)

Thanks for your thoughts and feedback. We agree with your points regarding the need to simulate different power system phenomena, including AC and DC networks, time and frequency domain simulators, transients and phasors, EMTP type and large power bus simulators for high numbers of nodes, detailed power electronics models, finite element methods, electromagnetic, electromechanical and thermal phenomenon, battery and solar cell physics, and so on. It's clear that no single simulation tool provides all of the functionality needed to meet the all the needs of the dynamic, fast changing electric industry.

 

We like to say internally that we try not to be religious about what simulation tools we endorse and support. Ideally, we would like to support all of them-- both for desktop co-simulation purposes and real-time hardware-in-the-loop purposes. Along these lines, one of the toolkits I should have mentioned above is the NI LabVIEW Model Interface Toolkit, which provides interfaces to more than 15 different simulation environments. (It is included with our real-time test and HIL environment, NI Veristand.)

 

You may also be interested in the recent press release by OpalRT regarding running their FPGA eHS Solver on CompactRIO hardware and new academic curriculum that spans from basic DC-DC converters up to three phase two level and multilevel NPC inverters. The OpalRT eHS Solver enables the conversion to real-time FPGA based simulations of models from Plecs™, SimPowerSystems™ (a Simulink™ product developed by MathWorks™), PSIM™, and NI Multisim.

 

Also, Hydro Quebec recently released a new timestamped solver for SimPowerSystems™ models executing on NI PXI Real-Time processors, which enables much more accurate power electronics simulations on processors (rather than FPGAs) by using the FPGA to accurately timestamp the switching events and then using that information in the real-time simulation solver running on an LabVIEW Real-Time RTOS processor.

 

My colleagues and I think it should be just a click of the button to move a simulation model from a desktop simulation to a real-time simulation on either an RTOS or FPGA hardware. (The emergence of good floating point capabilities and hybrid FPGA+DSP+uP system on chip devices is making the FPGA hardware accelerated targeting very feasible.)

 

The elephant in the room problem here is that every simulation tool vendor has their own proprietary format for the models, their own proprietary format for co-simulation interface, and their own proprietary format for real-time simulation (if any). This makes it very expensive for companies like NI to support all of the simulation model formats our customers would like.

 

On that front, we are very excited about increasing support for a standardized interface for simulation models called Functional Mock-Up Interface (FMI). This provides a standard, open, vendor agnostic XML based format for model description and model exchange. FMI is currently supported by 67 tools and the standard is advanced through the participation of 16 companies and research institutes.

 

To show our support, last month we released a new LabVIEW tool for FMI that enables desktop and LabVIEW real-time execution of FMI models using both the fixed and variable timestep solvers in the LabVIEW Control Design & Simulation Module. We would love it if you downloaded the tool, used it, and gave us some feedback good or bad. Right now it’s a free download on NILabs but we’d like to see enough customer interest and enthusiasm to justify fully integrating FMI into LabVIEW Real-Time and LabVIEW FPGA.

 

Another trend we’re excited about is Modelica. Modelica is a modeling language with multi-vendor and open source simulator support, and a large community of companies and researchers developing model libraries. (Developing good trusted models is actually a lot more expensive than creating a modeling tool/environment). Check out the open source PowerSystems library for Modelica.

 

OPENMODELICA is an open-source Modelica-based modeling and simulation environment that shows a great deal of promise, and we hope the open source community will rally behind it (and in particular that it will have best in class Functional Mock-Up Interface (FMI) support).

 

Everyone please let us know your thoughts. Do you agree with our philosophy and strategies to help address these issues? Where should we invest?

Message 3 of 4
(5,996 Views)

I think it's pretty cool that NI is thinking along these lines. Interoperability between modelling, simulation and software development tools is crucial.

I'm specially interested in FMI and Modelica. I've been studying the latter for a while. If I have the chance to play around with them in conjunction with LabVIEW I'll let you know how I find it.

0 Kudos
Message 4 of 4
(5,996 Views)