Power electronics converter control

[murgui]

Hello everybody!

I'm a electrical engineering student developing my final project. I've designed an inverter and a DCDC. I'm currently looking what hardware would let me implement the control designed. I don't know whether I'll implement the inverter or the DCDC.

The system has been designed with PSIM. The control is a PQ control in the inverter and a lineal & hysteresis in the DCDC.

The main worries in these controls are that the PQ includes a constant change between DQ-ABC axis which is a matrix product and also frequent product in the PI regulators. The DCDC includes many products in its PI regulators.

I have available a PIC24FJ64GA002 in a DM300028 developer kit. I mailed microchip wondering if any newer PIC would fit my developer kit.

My main worries are the availability of:

-3 or 4 PWM
-ADCs
-computing power

My PIC has 8 kB of RAM according to the datasheet, I don't know if this amount of RAM will be enough. The digital filters and the regulators might consume much RAM.

How do I should calculate the computing power and RAM necessary for this digital control ?

Any advice will be appreciated.
The image below is a good summary of the tasks to be performed. C&P transform, add, subtract and some products.



Thank you very much,

Murgui.

[Ttelmah]

This is all down to how you write your code....

For example, I've done twin channel PID with a loop update rate over 10KHz, with self 'learning' correction factors, on a PIC far less sophisticated (and slower) than this, using less than 500bytes of RAM. However I designed the maths to all use scaled integers treated as if they were 16bit decimal integers with an 8 bit decimal part.
Conversely you could waste all the chips performance, and use all the RAM in a few moments, by coding using float maths, and not understanding the implications of the code....
So (for instance) angle calculations, could potentially be done using a look up table, and solved in a fraction of the time, and with less RAM, than is involved in doing it using standard trig functions. However this comes down to you spending the time to understand the processor, and look at your loop mathematics and work out what accuracies are needed, and which components in the maths have the highest time overhead.
An enormous amount depends on what loop update rate is required by your maths?.

[murgui]

I understand that it's a very open problem. Do you know any specific reference concerning this kind of optimization?

Murgui

[Ttelmah]

Some of the MicroChip examples use similar code for simple PID. It was the 'thoughts' in these that I based my original designs on.