Tuesday 21 December 2010

End of term supervisor meeting

Recently I've been:
  • Building up a general knowledge of domestic appliances' power and energy characteristics
  • Getting used to using matlab to manipulate and display data
  • Plotting data from Gopal and Enrico's houses to understand trends in real data
  • Reading the Bishop machine learning text book. I'm currently reading the probability chapter
Over the holiday I'm planning to
  • Read more of the Bishop machine learning text book
  • Make some small UI changes to the GridCarbon Android app
  • Mark some HCI courseworks for Mark
When I get back from the holiday I'm planning to:
  • Look at techniques from other domains that can be used to recognise appliance's repeating power signatures, starting with the fridge

Monday 20 December 2010

Comparison of Gopal and Enrico's data

From the experiments I carried out last week, it became clear that houses with different occupants and appliances generate hugely different data. A problem with this is that some techniques will be useful on one data set and useless on another. To illustrate this, I've plotted the changes in power in both Gopal and Enrico's houses over a 1 week period. Increases in power are represented by red points while decreases in power are represented by blue points.

Gopal:


Enrico:


There are some similarities between these graphs in the density of data points. Each graph has a dense band of power changes between 0 and 300W. This band is visible 24 hours a day so it is most likely due to continuously cycling appliances, e.g. fridge, freezer, heating. Unfortunately, these cycling appliances do not have consistent 'on' and 'off' jumps and might therefore disguise the features of other low power appliances, e.g. TV, computer, lighting. Above this band (300-3000W), the data points are more sparsely distributed. These are likely to correspond to appliances that generate a large amount of heat, e.g. kettle, toaster, microwave.

However, there are also some differences between the two plots. Enrico's graph has a third visible band, between 0 and 20W. This is likely due to a cycling appliance with a consistent 'on' and 'off' change in power.

Another clear difference between the plots is the time of day at which the sparsely distributed data points occur. On Gopal's plot they occur throughout the night, whereas on Enrico's they only occur during the day. This is because Gopal has an electric boiler which runs during the night, and it is clear that this is hard to distinguish from other high power appliances.

Friday 17 December 2010

Real data from Enrico's home

Here's another plot of the changes in power from Enrico's home. Enrico has the largest amount of uninterrupted energy data, and I've therefore been able to plot 5 months worth of data (July - November). Enrico's data is was again collected from AlertMe.


This plot is really interesting in that there's a solid baseline of approximately 200W throughout the day and night. This is most likely because there is some appliance(s) that continuously switch 'on' and 'off' throughout the day. In addition, the jump in power ranges between 0W and 200W.

There's also two clusters representing morning and evening when appliance 'on' and 'off' events are more likely to occur.

I've also plotted the raw power values at minute intervals below.


In addition to the trends visible on the first graph, this also shows the morning and evening clusters separate from the baseline.

Wednesday 15 December 2010

Real data from Gopal's home

Here's a similar plot of changes in power, collected at minute intervals over a 1 month period. The system used this time was AlertMe. This plots when a change in power occurs, against the time of day.


This plot really doesn't seem to provide much information, as there aren't distinct clusters like were visible in Alex's data. I'm not sure why this is, but it could be due to an appliance with unpredictable 'start' and 'stop' signatures.

However, it's also interesting to just plot the raw data (power values against time of day).


Surprisingly, this shows a number of distinct clusters. While the first plot contained a data point for each 'on' and 'off' event, this graph contains a data point for every minutely power reading. This shows a clear repeating pattern each day, and although it isn't so useful in identifying which appliances are 'on' at certain points, could be used to predict when already identified appliances are likely to be used.

Tuesday 14 December 2010

Real data from Alex's home

Today I've been looking at some real data collected over a 4 day period from Alex's home. A large proportion of my time has been spent getting used to matlab, but I'm getting there slowly. I extracted changes greater than 100W in his overall power demand over the period. However, this is a 1-dimensional feature vector and does not make a very interesting plot, so I've plotted it against the time of day at which it occurred. The plot is below.


Any points that sit on a horizontal line are power jumps of the same value, and are therefore likely to be the same appliance. A few of these lines are visible over this small data set which is encouraging. The clearest of which is near the bottom left corner.

In addition, there's a clear cluster just to the right of the centre of the graph. This shows that time of day is a more useful feature vector for some appliances than others.

Appliance survey

Yesterday I spent the day building up my general knowledge of domestic appliances. I looked up ballpark figures for each appliance's average power during use. This demonstrates which appliances are likely to make characteristic step changes in power when they turn on and off. The results are shown below.

There's a clear split between appliances that generate heat (right side) and those that do not (left side).

However, we care more about appliances that consume most energy, not that draw the most power. To compare the appliances' energy consumption, I needed to estimate their duration of usage each day. I assumed they were present in a standard house of 2 people. The results are below.

It's import to note that I've excluded the electric boiler to aid comparison, as its consumption dwarfs that of other appliances. We now see things like lights creeping up the scale, so these are things we should care about more.

It's also worth noting what the power drawn by these appliances looks like. Many appliances change state while in use, and therefore draw varying power. These changes of state often repeat, and I've therefore classed these appliances according to whether their power demand is cyclic or constant. The fridge, washing machine, dishwasher and clothes drier are all cyclic appliances, and the rest are constant.

Monday 13 December 2010

End of week 9 and week 10 supervisor meeting

Recently I've been:
  • Finishing the introductory material to the Bishop machine learning text book
  • Using matlab toolboxes to experiment with techniques for event detection and feature extraction
This week I'm going to:
  • Build up a better general knowledge of domestic appliance power signatures, e.g. how much power appliances draw, average energy, cycle of signature etc
  • Spend more time than last week to get through a bigger chunk of the machine learning book

Monday 6 December 2010

Week 9 supervisor meeting

The outcomes of the meeting were:
  • Last week I wrote an introduction skeleton for which the structure and content was fine. I've tweaked and expanded some sections where recommended.
  • I've been making a grid of approaches in the literature to aid their comparison. So far I've compared literature from 2010 and 2009, but am going to continue populating this grid over the next week.
  • One of the common tools used in the literature is the Neural Network approach. Although I have a basic appreciation of the topic, I'm going to need to read up on the underlying theory in order to evaluate the benefits of various approaches.
  • With regard to data and experiments, I'm now at a stage where I have data from 3 houses, collected from both AlertMe and Ploggs available to me. This week I'm planning to use Matlab to apply some basic techniques from the literature to our data.

Wednesday 1 December 2010

November 2010 Activity Log

This month I had planned to:
Identify and fill any gaps in literature survey on NIALM.
Thoroughly read most relevant papers in the subject of NIALM.
Read a good chunk of the Pattern Recognition and Machine Learning text book.
Collect a range of research in the HCI/energy field.
Get hold of some real large scale aggregated appliance data from both Ploggs and AlertMe meters in a residential setting.

This month I achieved:
Collection and reading of a good coverage of NIALM literature.
Comparison of literature using conceptual map and grid of techniques.
Read the probability theory section of the Machine Learning text book.
Got hold of 1 week of real domestic energy data and performed basic analysis using Excel
Investigated Gaussian Processes.

Next month I plan to:
Collect data from a variety of households, metered using different techniques.
Use Matlab to apply increasingly complex machine learning techniques described in literature.
Read up on relevant parts of HCI and Ubiquitous computing fields

Research Methodologies - Introduction exercise

Issue

A Non-Intrusive Appliance Load Monitor (NIALM) refers to a device that can determine the operational state of connected electrical appliances from a single point of measurement. It aims to provide the advantages of sub-monitoring approaches without incurring the cost or intrusiveness of installation.

Importance

The output from a NIALM can empower the consumer by providing a breakdown of how much electrical energy individual appliances use. This can directly highlight energy savings through change in lifestyle or appliance replacement. Additionally, it can provide the input necessary should an agent seek to control appliances in the home in an energy efficient manner. Ultimately, this small-scale reduction in energy use contributes towards the global challenge of minimising energy generation and greenhouse-gas emissions.

Solutions

Previous solutions to the field of NIALM utilise high-granularity electrical readings, typically collected using custom-built meter. Approaches generally use signal processing and machine learning techniques to disaggregate the combined readings to give a break-down by appliance.

Gap

However, there is little research on whether such approaches can be applied to low-granularity data. Such low-granularity data will be measured by modern smart meters, which will become ubiquitous within UK homes by 2020.

Contribution

This research focuses on the application of NIALM techniques to low-granularity data that could be collected from smart meters. It aims to enrich machine learning NIALM techniques using additional data sources such as collaborative appliance models, and temporal, user and environmental data.

Order

The order of this work is as follows. First, the goals and applications will be stated. An in depth discussion of related literature and its limitations will then be given, followed by a description of our contribution. Our proposed approach will then be evaluated against existing approaches before the conclusions are highlighted and future work is suggested.

Monday 29 November 2010

Top 6 lists for Research Methodologies Course

Top 6 people:
G. W. Hart
M. E. Berges
M. Baranki
L. Farinaccio
S. B. Leeb
L. K. Norford

Top 6 conferences:
Ubiquitous Computing, ACM International Conference, ACM UbiComp
Autonomous Agents and Multi-Agent Systems, AAMAS
Computer Supported Cooperative Work in Design, IEEE CSCWD
Instrumentation and Measurement Technonlogy conference, IEEE IMTC
Sensor Mesh and Ad Hoc Communications and Networks, IEEE SECON
Intelligent Data Engineering and Automated Learning, Springer IDEAL

Top 6 journals:
Computer Applications in Power, IEEE
Energy and Buildings, Elsevier
Instrumentation and Measurement, IEEE Transactions on
Autonomous Agents and Multi-Agents Systems, Springer
Machine Learning, Springer
Electrical Power & Energy Systems, Elsevier

Friday 26 November 2010

End of week 8

This week I have concentrated on collecting as much of the NIALM literature as possible. This has not been a trivial task, as few papers appear in the same journal or conference, however I feel I have a good coverage of the field now. I've been reading the most relevant ones and making notes on the techniques they employ. One promising theme rising out of the most recent papers is rising interest in the low granularity data that could be obtained from smart meters.

My plan for next week is to:
  1. Fill in my grid comparing techniques used in each paper
  2. Do research methodologies tasks (find top people, conferences, journals, reflect on key conference paper's abstract and intro)
  3. Use matlab to apply some techniques from the literature

Thursday 25 November 2010

Post-meeting direction

I'm going to focus on what we can learn from data that can be could be collected from smart meters. Data of granularity this course is rarely used in the literature, as the accuracy of inferences are likely to be lower than for higher granularity.

My to do list:
  1. Continue literature search by following references of most recent papers
  2. Continue to read most relevant papers
  3. Make a grid to compare techniques used in each paper
  4. Do research methodologies tasks (find top people, conferences, journals, reflect on key conference paper's abstract and intro)
  5. Use matlab to apply some techniques from the literature

Monday 22 November 2010

In-depth literature search

Following my library skills session this morning, I conducted a more thorough search of the my field of NIALM.

I searched the following databases:
  • Web of Science
  • Inspec
  • Compendex
  • LNCS
For the following search terms:
  • Non-Intrusive Appliance Load Monitoring (relevant results)
  • NIALM (few, very relevant results)
  • NILM (non-domestic applications of load monitoring)
  • NALM (many irrelevant results)
Among the masses of papers I collected, the following are the most interesting:
  • Ruzzelli AG, Nicolas C, Schoofs A, O’Hare GMP. Real-Time Recognition and Profiling of Appliances through a Single Electricity Sensor. In: Sensor Mesh and Ad Hoc Communications and Networks (SECON), 2010 7th Annual IEEE Communications Society Conference on.; 2010:1-9.
  • Murata H, Onoda T. Applying Kernel Based Subspace Classification to a Non-intrusive Monitoring for Household Electric Appliances. In: Dorffner G, Bischof H, Hornik K, eds. Artificial Neural Networks - ICANN 2001.Vol 2130. Springer Berlin / Heidelberg; 2001:692-698. Available at: http://dx.doi.org/10.1007/3-540-44668-0_96.
  • Benyoucef D, Bier T, Klein P. Planning of energy production and management of energy resources with Smart Meters. In: Advances in Energy Engineering (ICAEE), 2010 International Conference on.; 2010:170-173.
  • Pihala H. Non-intrusive appliance load monitoring system based on a modern kWh-meter. Vtt Publications 365. 1998;(May).
I'm planning to start reading through each of these four tomorrow morning, in order of relevance.

Library skills

I completed the ECS library skills for research postgrads this morning. It was a session of just over an hour long, to inform us about the different library services we can use, and the electronic subscriptions the university holds.

Although the session mostly covered material we'd previous been taught, the main benefit was being asked to perform a literature survey using the sources they recommend. Even though the sites and databases were again nothing new, the exercise of investigating the literature without being distracted by reading specific papers was really helpful. I'm going to conduct a similar, but more in-depth search this afternoon.

Friday 19 November 2010

End of week 7

This week I've spent my time working on two topics. The first has been playing with Gopal's electricity readings and the second has been reading up on machine learning theory.

First I needed to get hold of Gopal's meter readings to the highest granularity possible. I've written a Python script which downloads power readings for a given time period at one second intervals. Using that script, I downloaded a week's worth of data. After plotting the data and having a few conversations with Gopal about his appliance use, we made a few observations:
  • The data contained a bunch of erroneous readings, which were represented as 0s in the data.
  • He has a time-of-use electrical boiler, which heats up a tank of water at midnight and maintains this temperature by periodically switching the boiler on until 8am.
  • His absolute minimum demand is 70 W. We think the 70 W is made up of his Sky box (35 W), a standby TV (10 W) a wireless router (10 W), an extractor fan (5 W) and some other uncountable appliances (10 W).
  • However, his average 'always-on' demand is roughly 370 W, his fridge-freezer accounting for the rest. The power drawn by the fridge-freezer seems to vary quite a lot, with a pattern which is does not visibly repeat.
  • The energy values given by AlertMe are very similar to the integral of the power demand over time. However, there seems to be a difference of between 0 and 1% between the values that I calculated and the values that AlertMe return.

The other thing I've been up to is reading lots of theory about machine learning. I've read the introduction to probability theory in the Bishop book. I also asked Ruben how to find out more about Gaussian Processes, and he pointed me at chapters 2 and 3 of this book. I'm finding it pretty heavy going, but at least understand what they are now, and am hoping to really get stuck into their applications next week.

Wednesday 17 November 2010

A better way to visualise power data

I've found visualising power data on a scatter plot is more useful than on a line graph:



Without the line interpolating between data points, we can clearly see the difference between continuous power demand and single data points which are part of a transient. I've also excluded 0 values. This is because 0 W demand is most likely an error, as no readings exist between 0 W and 70 W readings.

VirtualBox setup

The FigureEnergy project I worked on over the summer involved programming of a system to empower consumers in managing their energy consumption at home. A major part of this project is the annotation of energy usage with what appliances or activities were going on. This is clearly highly related to the HCI element of my PhD.

To enable me to make any changes to the FigureEnergy deployment, I need to use a python library called fabric. Fabric wasn't designed for use within a Windows environment, so Enrico and Gopal recommended I set up a Ubuntu VirtualBox.

After downloading VirtualBox and creating a new virtual machine and hard disk, I installed Ubuntu 10.10 from a disk image. The next step was to get at my Windows files stored on the host machine's drive. To do this I followed the steps in this tutorial, which involved installing guest additions, sharing the folder with the VirtualBox application and then mounting the folder within Ubuntu.

Once I'd done that, I needed to install fabric. Given that Ubuntu came with Python 2.6, all I needed to do was install pip using the Ubuntu Software Centre, and then install fabric using pip.

Monday 15 November 2010

Preliminary steady-state analysis

I've collected data for the total power demand of a home sampled at 1 second intervals for a 24 hour period. Here's the plot:




There are loads of peaks at around 2000 W, which are easy for a human to recognise. I wanted to pick these positive and negative step changes out automatically, so I wrote a script to calculate the difference between adjacent power readings, and if the difference was greater than 1000 W, record them. I've plotted the result:


I was hoping to see each red dot sitting roughly level with at least one blue dot and visa versa. This would mean that the step change for the appliance as it turns off is roughly the same as when it turns on. However, very few of red and blue dots actually sit on the same level, and even worse there's not even the same number of red as there are blue.

Tomorrow I'm planning to compare readings within a window, instead of using adjacent readings.

End of week 6

I started this week by revisiting Ford's thesis 'Reducing Domestic Energy Consumption Through Behaviour Modification'. I gave particular focus to the probabilistic model used to combine machine learning outputs with additional data, e.g. time of day and appliance average usage time.

I also read Mudasser's 9 month report before attending the first Research Methodologies course. The focus was on the general time scale of a PhD and the structure and content of the 9 month report. Specifically, I found the most useful outcome finding out about the flow of reports. We were told how the introduction should introduce each topic in the literature review. This would ensure that only relevant literature was discussed, and the review was therefore kept closely on-topic.

We also had a visit from Mike Osbourne, a researcher from Oxford University. Mike's interests lie in machine learning, specifically the application of gaussian processes. We discussed the how such techniques could be applied within the energy domain as part of the ORCHID project. We concluded that in the area of NIALM, there was a significant gap for this kind of approach.

Finally, we have started an Energy Special Interest Group within ECS. I have created a wiki, which the group is planning to use to share literature and ideas. I would encourage anyone within the group to visit it and expand it with your area of interest/research.

Thursday 4 November 2010

GridCarbon v1.1

I have just published an update for the GridCarbon Android app. This release fixes the layout bug present in the first release.

I only found this bug through the process of submitting the app to the Android Market. The Market requires all apps to define a minimum sdk version, which explicitly states which devices the app should be compatible with. For some reason, setting this flag caused much of the content to layout in a slightly ugly fashion on devices with high resolution screens. I fixed it by changing the dimensions of some elements from fixed pixel sizes to resolution-independent sizes.

Tuesday 2 November 2010

Aims for week 5

After my weekly meeting with my supervisors, this week I'm aiming to:
  1. Read (the most relevant sections of) a PhD thesis titled 'A Framework for Enabling Energy-Aware Facilities Through Minimally-Intrusive Approaches'
  2. Get hold of some AlertMe data and play with it to compare it to Plogg data
  3. Fix the layout issues of the GridCarbon Android application
  4. Sort out my ECS homepage so it describes my research

Monday 1 November 2010

End of week 4

This week I've been focussing on learning the basics of pattern recognition and machine learning. I've borrowed a copy of a text book and read the opening chapter, and Alex has ordered me a second copy so I can really get stuck in to it.

After our meeting last week, I had a closer look at what data Ploggs offer, the full details are in the previous post. I looked at energy readings vs power readings, and also at frequency of readings. Next week I'm going to look in to the data available from AlertMe meters and compare it with Plogg data.

I've been in contact with Mario Bergés, a researcher at Carnegie Mellon University who is working in a very similar field. He's got some very interesting papers in the pipeline, and has also kindly sent me a copy of his PhD thesis.

Towards the end of the week, I've been putting some more work into GridCarbon, an Android application to monitor and visualise the carbon internsity of the UK electricity grid. I've prepared and published the first version of this application to the Android Market, although v1.0 has suffered from layout bugs due to backwards compatibility issues. More specifically, the Android Market requires the application to specify a minimum API level. However, when this attribute is specified, the application's layout takes a hit.

Thursday 28 October 2010

Energy vs Power in NIALM

After having a play around with some real data from a Plogg connected to my machine, I began to wonder, which is more useful, energy consumption or power demand recorded from my Plogg? I plotted both over the last week:





After a glance at these two graphs, it becomes apparent that energy is not a function of power over time. In fact, a reading of the power demand at an instant in time is recorded at each interval. However, all the energy used during the interval period is accounted for from the energy readings.

These properties have both advantages and disadvantages for appliance monitoring. A major disadvantage of recording power values at time instants is that activities not taking place at the time of the reading are missed. An example of this is the spike clearly visible on the energy graph, but not recorded on the power graph.

However, the power graph does have an advantage over the energy graph. Because no averaging is done over the time interval, the reading reflects the exact sum of the appliances' power demands. This makes NIALM an almost trivial task. This is not quite so easy on the energy graph, as changes in appliance state between readings are likely and an average of both states is therefore recorded by the following reading.

Monday 25 October 2010

Topic discussion

I had a chat with Alex today about my topic and what direction to read into within the literature. Here's what we decided:
  • Most approaches in the literature focus upon highly detailed data from expensive metering equipment
  • We want to look at what we can find from less detailed detailed data, captured from inexpensive monitoring equipment
  • The characteristic we want to measure is energy (or power), not current, voltage etc.
  • The granularity of data that should be available is in the frequency range of one reading per second
  • We want to investigate how complete the coverage of appliances needs to be. Some expensive equipment claim to recognise 99% of appliances. Is 90% just as useful though?
  • We also want to investigate when it is most important to prompt the user to tag an unrecognised appliance
This week I'm going to focus on the machine learning fundamentals, which will ultimately be used to extract features from the data, and match these features to known signatures.

Friday 22 October 2010

End of week 3

This week I have read a bunch more papers on NIALM, and collected five of the best papers which provide an overview of the topic. These overview papers are few and far between, as the vast majority of papers contain a tiny background section and an in depth description of a proposed technique.

I've also been to a few more lectures on sustainable energy and power systems analysis. The lectures from both modules seem quite interesting and I'm working through some exercises with Sam.

I have been really impressed with Mendeley reference manager application, and am now using it to share references with Sam and also to store my reference database in the cloud.

I've continued to work through the Intelligent Agents course material, with only two lecture's worth left. I hope to finish it early next week and move on to look at some machine learning material, and how it can be applied to NIALM.

I had a chat with Bhargav about his topic for his MSc project. As I understood it, his project is focusing on the application of machine learning techniques to data collected from sub-metered appliances, with the aim to provide automatic behavioural interventions to encourage energy savings. We discussed how research detailed heavily in the literature discusses modelling appliances as Finite State Machines, and how this high-level information could be useful in providing behavioural interventions.

Inspired by this thought, I wondered whether this would be apparent in my own data collected by a Plogg connected to my work computer and monitors. With a quick plot of the half-hourly power and energy consumption I could already see a number of different states from the power saving settings of my machine; when my computer was on, when it had turned my monitors off, when it had gone to sleep, and when it was off. The fact that these states are recognisable at the low-grained data of 30 minutes intervals of the load data has convinced me that it should be possible to construct an FSM for a specific computer, or even a generalised FSM for a number of computers.

Wednesday 20 October 2010

Top papers for Non-Intrusive Appliance Load Monitoring (NIALM)

Update 09.09.2011 - I have since posted an updated list of NIALM papers.

Below are some of the most useful overview papers for NIALM.

1. Hart GW. Nonintrusive appliance load monitoring. Proceedings of the IEEE. 1992;80(12):1870-1891.

Comprehensive description of problem area and review of work to date. Although the approach implemented is now considered basic by today's standards, the outline of possible solutions is still applicable today.

2. Najmeddine H, El Khamlichi Drissi K, Pasquier C, et al. State of art on load monitoring methods. In: Power and Energy Conference, 2008. PECon 2008. IEEE 2nd International.; 2008:1256-1258.

A concise and up-to-date overview of the groups of approaches towards NIALM. Although short, this paper is useful because its focus is many existing techniques, as opposed to one newly proposed technique.

3. Laughman C, Lee K, Cox R, et al. Power signature analysis. Power and Energy Magazine, IEEE. 2003;1(2):56-63.

Detailed comparison of a number of NIALM approaches, with a discussion of their relative advantages and disadvantages. The effectiveness of the techniques within different setting, e.g. domestic, commercial, industrial is also evaluated. This paper is useful for the same reason as the last; the focus in on the range of techniques, not one specific implementation.


4. Matthews HS, Soibelman L, Berges M, Goldman E. Automatically Disaggregating the Total Electrical Load in Residential Buildings: a Profile of the Required Solution. Proc. Intelligent Computing in Engineering.381-389.

In depth overview of the practicalities of approaches proposed in the literature. Describes not only research methods in this area but also problems with modern commercial metering.

5. Ting KH, Lucente M, Fung GSK, Lee WK, Hui SYR. A Taxonomy of Load Signatures for Single-Phase Electric Appliances. In: IEEE PESC (Power Electronics Specialist Conference).; 2005.

Short discussion of why appliances present distinct signatures, and a proposal of how these signatures can be grouped. Since pattern recognition is generally improved through prior knowledge of appliances, the idea of identifying certain signatures out-of-the-box is an attractive one. However, the signatures described use a sampling of current and voltage at 50Hz, a data granularity far greater than that available from most modern smart meters.

Monday 18 October 2010

End of week 2

I've now set up my machine and got it ready for work. Didn't have too many problems, though I did need an HDMI to DVI converter for my second monitor, and also had to change the boot order to stop it trying to boot from the network.

I've attended a bunch of lectures this week from the Energy Sustainability course. I've decided to carry on with the modules 'Power Systems Analysis' and 'Sustainable Energy Systems, Resources and Usage' for now, but drop 'Conventional Generation Technology' as it's not too relevant to my topic. It appears that the remaining module, 'Introduction to Energy Technologies', was a one off lecture and will not be continued throughout the semester.

I've conducted a more thorough literature review on the existing solutions to the Appliance Load Monitoring problem. Starting from the PhD thesis' references and working backwards, I've collected as many relevant papers as I could find. The topic was introduced by Hart, in the 1980s, and is reviewed thoroughly in his 1992 paper 'Nonintrusive Appliance Load Monitoring'. Since then, many algorithmic approaches have been suggested for automatic appliance disaggregation, largely consisting of both pattern recognition and signature clustering.

I've experimented with some reference manager tools with limited success, but am gradually building up a bibtex file of any resources I've found.

I had a meeting with Mark in which we mostly discussed the topic and methods of literature search. He advised I start to take a more formalised approach, as opposed to the random path through the literature I'd been following so far. He's also lent me a book 'Persuasive Technology' by B. J. Fogg, which will be very useful when considering the HCI aspect of my topic.

I've also decided to work through the notes for Alex's Intelligent Agents module, since I didn't take this in my 4th year.

I'm still reading 'Sustainable Energy - without the hot air', and am now just over halfway through.

Thursday 7 October 2010

First Week

This week has been the first week of my PhD, which has been mostly taken up my admin stuff. I've sorted out my computing account, got a desk and ordered a computer. I've also had a meeting with my supervisors, Mark and Alex, to define a topic, this was the outcome:

Investigation of collaborative human-agent appliance disambiguation from domestic energy use. Also, investigating the benefits of sharing the raw energy usage and derived recognition results over a social network.

I also attended one and a half days of compulsory induction material.

I am planning to attend some lectures from the MSc Energy Sustainability with Electrical Power Engineering, in addition to intensive courses on probability and game theory.

I have been reading parts of a PhD thesis titled 'Reducing Domestic Energy Consumption Through Behaviour Modification'. I've also been reading a couple of papers titled 'At the flick of a switch: Detecting and classifying unique electrical events on the residential power line' and 'Reducing domestic energy consumption: from psychology to technology'.

I have been reading the book 'Sustainable Energy - without the hot air' by David MacKay when I get the chance.