Saturday, September 1, 2012

PEPE - September 2012

After taking most of the last year off I have started back at work on PEPE. At this point I'm 50% refamiliarized with the algorithm but some of the code still looks like magic. I really should have left better notes, especially on the Synthetic Weather section of the algorithm. Meh... I'll figure it out eventually.

I'm working on two tracks at the moment. Track A) Go through the existing code, break it into blocks and move those blocks into modules. This process is rather mindnumbing but reorganizing the code is long overdue. Track B) Add a Load Simulation code.

I'm using the load model from "Domestic electricity use, A high-resolution energy demand model" as a guide. The load modelling procedure starts off with a simulation of active home occupancy - in this case Active Occupancy means you are home and you're also awake.  This section of code essentially looks at three things:

1. How many people live in the house?
2. What day of the week is it?
3. What time of day is it?

These three questions are used to look up a set of coordinates in a transition probability matrix. The transition probability matrices are built to take into account the behavior of normal people. For example, normal people are active during the day rather than at night. It's also normal for people stay up later on the weekends compared to the weekdays. The Active Occupancy model wasn't too hard to translate and reconfigure to my needs. I still have some tweaking to do but my version of the model seems to be working for the most part. As it turns out solving this Occupancy algorithm offered a smooth way to slide back into coding.

Anyways... once you have an Active Occupancy profile you can use the data to predict the likelihood of people using energy in the home. This makes sense when you think about the fact that people generally only watch TV and run the Microwave when they're home and awake. The goal here isn't to predict exactly when people are using electricity to watch TV - I'm interesting is predicting approximately when people are likely to use electricity. The overall goal is to build a bottom up load model. This bottom up model will compile the demand from individual loads - things like lighting, dishwashing, TV, stove.

Lights, Camera, Action

After the Occupancy model I moved on to my first load - Lights. I lifted the Lighting Model from the same jokers that put together the Occupancy model. These guys designed their lighting demand model around the rather reasonable idea that people are more likely to have their lights on if it's dark outside. The model came with a database that had generic sunshine profiles for 12 different days - one standard day for each month. If you wanted to know how sunny it was you'd look up the sunsine for that time of day and particular month in a data table. You'd then scale the table's sunshine value by a random number to come up with an Irradiance value. This lighting simulation method is all well and good for a quick and dirty Irradiance estimator but PEPE already has an hourly synthetic weather model which calculates Sunshine. It wasn't too difficult to fix things so that the Lighting Model is working off my Synthetic Sunshine data versus the table lookup method. At this point I still have another 10% of the Lighting Model to finish up but I think I have my head wrapped around what needs to be done. Tick and Tack... Can I get a Toe?

I've got a Sunday shift tomorrow which should be quiet enough to allow me to complete the Lighting Model and move over to the Appliance model. I'm thinking the Appliance model is similar to the Lighting model so it's shouldn't be all that hard to rejigger. Once the Appliance model is out of the way I'm hoping to move on to a water heating demand model and finally a space heating demand model.

All these individual energy simulations put together will give a simulated load profile for an idividual home. The load profile with be a synthetic simulation but if done right the simulation will be statistically accurate.

Once I have a simulated load profile the next step is to develop an energy management system strategy. The goal of the EMS strategy will be to maximize the home's solar self-consumption. This is done by identifying the shiftable loads and developing a shifting strategy. As a general rule shiftable loads have one of two qualities - A shiftable load either stores heat and/or the load performs some sort of flexible batch process like cleaning dishes or drying a load of laundry. Shiftable loads are sub-classified as Autonomous or Semi-Autonomous. Here's how the classifications of shiftable loads break down.

Refrigerator - stores heat and batch process (ice making), autonomous
Dishwasher - potentially flexible batch process, semi-autonomous
Washing machine - potentially flexible batch process, semi-autonomous
Dryer, - potentially flexible batch process, semi-autonomous
Water heater - stores heat, autonomous
Space heater - stores heat, autonomous
Pool pumps - potentially flexible batch process, autonomous
Air dehumidifier - potentially flexible batch process, autonomous

A refrigerator is a basic autonomous load. Here's how a refrigerator might operate when under the control of an Energy Management System

Scenario 1.
Fridge: Hey, EM (EM is short for Energy Management System)
EMS: What can I do for you?
Fridge: I'm currently running at X degrees.
EMS: I have no surplus power. Standby. Run down to X + 4 degrees. I should have excess power for you then. If not run a half-cool cycle back to X degrees.
Fridge: OK.

Scenario 2.
Fridge: Hey, EM
EMS: What can I do for you?
Fridge: I'm currently running at X degrees.
EMS: I have surplus power available. Run to lowest temperature (X - 6 degrees). That should take 22 minutes.
Fridge: Why are you so bossy?
EMS: Shut up and do your job.
Fridge: You're a dick.

Scenario 3.
Fridge: Hey, EM
EMS: What can I do for you?
Fridge: I'm running at 75% ice capacity. Is now a good time to top up?
EMS: I have surplus power available. Top up to 100% ice capacity.
Fridge: OK. Sorry about calling you a dick.

A dishwasher is a semi-autonomous load. By semi-autonomous I mean a user has to trigger a start but once triggered the machine runs through a pre-programmed cycle.

The Sitch...

1. You've got a load of dishes that you want done.
2. You don't need the dishes clean for a few hours so you press the Delayed Start Button
3. It's 8 a.m and the Sun is low on the horizon.

DW: Hey EM.
EMS: What can I do for you DW?
DW: I just got a Delayed Start command. What should I do?
EMS: I don't currently have any surplus power but the weather forecast indicates clear skies today. I should have enough power for you to run in approximately 1 hour.
DW: Sounds good. Slot me in for 9 a.m.

The idea behind all this load management is to maximize the amount of power delivered from the photoelectric system to the loads in the home. If we assume the solar power is cheaper than utility electricity it means that each extra sliver of solar power that is used saves the end-user money. Make sense? Good...

No comments:

Post a Comment