Development platform for a Home Building Energy Management System

Thread Starter

cyberwalker

Joined Feb 23, 2010
3
Can anyone please suggest a good development platform for research on a Home Building Energy Management System (HBEMS) project?+

The project involves :
Power Flow control algorithms

Artificial Neural Networks (hopefully adaptive) and

Time Scheduling with Genetic algorithms.

Do you have any suggestions?!
 

wayneh

Joined Sep 9, 2010
17,496
My suggestion is to leverage the work of others as much as possible. I mean, you could write your own neural network in C or just about any other language, but there's no reason to. Find what platform has most of the pieces you need already made. See what others in the field have done.
 

Thread Starter

cyberwalker

Joined Feb 23, 2010
3
My suggestion is to leverage the work of others as much as possible. I mean, you could write your own neural network in C or just about any other language, but there's no reason to. Find what platform has most of the pieces you need already made. See what others in the field have done.
Yes, exactly as you say. I am asking for ::
1. what platform has most of the pieces you need already made
2. what others in the field have done
 
Before you get to clever you need to think about connected...
retrofitting an old property is expensive and messy if you need to install cables.

You will end up thinking X10 or better still some form of wireless mesh network, ZigBee esque.

The reason I mention it is that the choice of network (PAN), unless you start from scratch with your own new design, will dramatically affect how you go about controlling things.

Bear in mind that the cleverer you make your devices the more they will cost and you will have to be exchanging exponentially more data which brings you back to basic topology.

I would humbly suggest that before you look at the detail you plan some concepts. Don't get me wrong I think its a great idea but then so are humanoid robots with SiFi IQ's.

If you want to get it off the ground as a project you need to look at the data you can realistically collect and what, if any, benefit an adaptable system could derive from that.

To give you a hard example ...
Many hotels now have there heating systems zoned by room and linked to the booking control. The room knows when it is likely to be occupied and starts to warm up a little. When it gets occupied, pir's and door sensors and the like, it ramps up to its ideal occupancy temperature. O and if you open the window and leave it that way it will assume you are too hot and turn down the heating.
It might also know what time you set the alarm for in the morning ...

On top of that since rooms can to some extent predict what their energy requirement is likely to be they can tell the centralised heating and hot water systems when demand is going to be high or low.
A big hotel can save a fortune by only storing enough hot water for what is likely to be required and if rooms are intelligently booked in blocks or zones, when occupancy is low, whole sections of the building can go into low energy standby with flow temperatures and rates reduced.
 
Last edited:

Thread Starter

cyberwalker

Joined Feb 23, 2010
3
@Dyslexicbloke, this (and many more) is what my system does already. And in models it succeeds. It's an academic project which may stand a chance to be a product. I have done a lot of research on it, have a decent background in the science, all I need is a good suggestion for a development board for a HBEMS. Thank you very much for your time.
 
Mmmm "And in models it succeeds" ... Well ...

Assuming you have managed all the complex intelligent stuff but somehow failed to come across any mundane reality I will try and point out a few real world issues that you will need to address when talking specifics.

You should probably be looking at a wireless mesh system, at least for the processing nodes, data collection points and centralised management are another matter.

Quite apart from the processing you need for your base application you will need to manage a wireless network, a PAN, intelligently, lets assume it is ZigBee as I suggested above, although there will be many other choices.
A commercial product that springs to mind is Spinwave, a packaged PAN mesh supporting proprietary devices as well as serial coms that can communicate with just about anything.

You need a microprocessor in each node that has serial communications hardware built in.
Preferably you want a single chip solution, something fairly quick with plenty of headroom.

Again many to choose from but quick wins would be Atmel, Parallax (Basic stamp) or PIC, and there many respective guises. That way you get a relatively straight forward development environment but on a product/platform that could easily scale to a retail device, with many of the the bells and whistles omitted.

Now you need sensors, again we are talking wireless because of the issues with masses of cabling. You probably also want devices that scavenge environmental energy and only transmit data when it is pertinent

Try googling ' Energy Harvesting Wireless Sensor ' there are over 1000000 hits that should give you some specifics.

Lastly your ultra low power sensors will need to be node clients, as opposed to functional PAN members, and if your sensor density is high you will need to manage channels and frequencies as well as keeping them well outside the PAN range so will probably need a completely different wireless system for those.

The point is you don't need 'a' platform you need at least 3 totally different ones almost certainly all of which will need to interface with currently available products.

When you did your successfully modelling did it include real world communication and power requirements, I suspect not.

I am just a self taught engineer, no further education of any sort but I can assurer you that until you identify, or design and build, your component parts you cant possibly expect to make an intelligent decision with respect to platforms for retail products, which are seldom the same as development environments.

The requirements for a sensor are likely to be worlds apart from those that define a PAN node which will in turn will be completely different from a system management hub and data processing centre.

If your serious 'do the work' and then ask questions targeted at 'real' problems because unless you are planning to set up in silicone valley and take the world by storm with a new family of electronic devices you will need to work with what is available, even if your implementation is a new and innovative use of existing technology.

I wish you the best of luck but fear you may be attempting to miss out the required fundamental design process.

However intelligent and adaptable your algorithms are they will be useless if they are deaf, dumb, blind and unsustainable.

Academic projects are a hugely impotent basis for real world applications, but it generally takes far longer with much more effort to realise the dream. Look at any pinnacle technology, on any scale!
 
Last edited:
Top