Modelling problem

I think I am in good track to mathematically model the energy problem for 9-month report using multi-agent systems.

Typically, one of the solution to reduce energy costs and carbon emission is to flatten the energy consumption as much as possible. Given this, my model generally has three main things to sum up:

- Tasks allocation (agent can schedule  the list of activities on the specific time scale for consumers)

- Human-Agent Interaction (Users can interact with the scheduling)

- Agent can learn activities in the past to have a better schedule in the future.

This model also takes FigureEnergy into account as the system requires users to annotate their activities daily.

I am writing up my models soon after having discussion with my supervisors :).

FigureEnergy study preparation (draft version)

FigureEnergy study preparation.

1 – Purposes:

By using FigureEnergy (FE) system, we want to test our assumption that there is a difference between using the annotation system and not using the annotation system in the weekly energy usage changes and in the energy knowledge gained after using the FE system. We apply two reward mechanisms to encourage participants during the experiments, and we want to see which reward mechanism is better.

2 – Rewards mechanism

We define:

-           “Flat” reward: is to give participants a fix amount of money x if they manage to save y% compared to their prior usage

-          “Linear” reward: is to give participants £1.00 per energy N KWh they save

3 - Measures

-          Weekly/Daily energy usage

-          How much knowledge the participants gain from their energy consumption changes. This is required participants to complete two questionnaires, which are “before the experiment” and “after the experiment”

4 – Participant’s conditions

-          Commitment of using AlertMe non-stop during the experiment (i.e. not turning off AlertMe’s hub power, internet...)

-          The main cable of the participant’s house is not too far away from the Ethernet. The maximum range between the hub and the Meter Reader is 20 metres and a signal can be detected through 2 walls. If the Signal Strength icon is flashing, it means that your Display has a weak connection with your Meter Reader. Therefore, you need to try to situate your Display closer to your Meter Reader.

-          Participants should have previous computer experience.

-          Participants should not take any vacation during the experiment

5 – Plans before experiments

-          A clear instruction about the experiments

-          A short information pack of energy consumption (1-2 pages) is to allow participants aware of energy saving. This approach can increase energy awareness to all participants and make a fairness

-          Each experiment would be held for 1-2 months

-          Two questionnaires need to be prepared for participants to do before and after the experiment. This way we can measure how much knowledge they gain after using the system.

6 – Experiments

Experiment 1:

In this experiment, we use a “flat” reward mechanism as a motivation key, which is to give participants a fix amount of money of £10.00 if they manage to save 10% compared to their prior weekly usage.

We assume that participants who use FigureEnergy system with the annotation feature will gain more energy knowledge and save more energy than the participants who use FE system without the annotation feature.

Experiment 2:

This experiment process is exactly similar to the experiment 1. We swap the “flat” reward mechanism to “linear” reward mechanism. The “linear” rewards mechanism is to give participants £1.00 per energy KWh they save, compared to their prior weekly usage.

-------------------------------------------------------------------------------------------------------------

When these experiments are finished, we can compare the results to make an insightful analysis.

Energy figures sent by AlertMe (from 08 Feb 2011 to 14 Feb 2011)

These data are looking so bad:

{'values': {'average': [9.0604011100000008, 9.0819194400000001, 9.1046013899999991, 9.1281238899999995, 9.1520869400000002, 9.1736197199999996, 9.2176438899999997, 9.2176438899999997, 9.2611055600000007, 9.2828147199999993, 9.2828147199999993, 9.3246222200000002, 9.3444677800000004, 9.3633580599999995, 9.3824405599999992, 9.4013088899999993, 9.4203130599999998, 9.4393769400000007, 9.4583880600000008, 9.4779338899999992, 9.4992638899999999, 9.5192144400000007, 9.5373694400000009, 9.5547061099999997, 9.5723149999999997, 9.5897733299999999, 9.6072930599999999, 9.6072930599999999, 9.6250344400000003, 9.6649991699999998, 9.6825041699999996, 9.6825041699999996, 9.7171136100000002, 9.7350327799999992, 9.7529280600000003, 9.81064194, 9.8285041700000004, 9.84877556, 9.8666927799999993, 9.8839755599999997, 9.9011197200000005, 9.9011197200000005, 9.9359955600000003, 9.9359955600000003, 9.9535327799999997, 9.9904713899999997, 10.010261699999999, 10.010261699999999, 10.0454261, 10.0454261, 10.0637308, 10.0637308, 10.1193683, 10.1193683, 10.1378389, None, 10.1378389, 10.1378389, 10.1378389, 10.1378389, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, 15.3055503, 15.3055503, 15.3562878, 15.3804047, 15.4045486, 15.4305447, None, 15.4305447, 15.4305447, 15.4305447, 15.4305447, 15.4305447, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, 18.390035600000001, 18.390035600000001, 18.390035600000001, 18.437413599999999, 18.455865299999999, 18.455865299999999, 18.493091400000001, 18.511278600000001, 18.5295606, 18.549485600000001, 18.568988900000001, 18.587255299999999, 18.605908100000001, 18.6244172, 18.642919200000001, 18.663818899999999, 18.685745799999999, 18.7028344, 18.7208714, 18.7208714, 18.7208714, None, 18.7208714, }

FigureEnergy graph display issue

I was quite excited to see my energy consumption at home after I installed the AlertMe mornitor 1 week ago. In addition, I can play some functions on the FE system to see how efficient that the system would offer.

Unfortunately things never come smoothly. I could not see my data displayed properly, and Gopal's data also went doubtfully crazy on some specific day (eg. 12 Feb 2011, that could be a "black" saturday)

I checked and tested the system yesterday for a while and discovered another extended issue of the "zero-gap". Basically the zero-datas not only appear at the end of the list of data, they also can be somewhere in the middle of the list. The previous algorithm couldn't recognise the "zero-gap" in the middle of the list. It considerer these data as the correct figures, then plot a graph with a very big value (e.g 4000KWh) while others figures were just about less than 1.00kWh). Therefore, this is why we could not see our energy consumption properly.

For example, here is the list of the data sent from AlertMe server.

{...
4278.4315699999997, 4278.4370399999998, 4278.4470700000002, 4278.4553900000001, 4278.4636700000001, 4278.4834700000001, 4278.4988199999998, 4278.5049799999997, 4278.51944, 4278.5346499999996, 4278.55033, 4278.5656600000002, 4278.5823700000001, 4278.5986599999997, 4278.6212800000003, 4278.6316999999999, 4278.6432000000004, 4278.6551799999997, 4278.6613900000002, 4278.6729100000002, 4278.6819299999997, 4278.68833, 4278.7007599999997, 4278.7067299999999, 4278.71425, 4278.7236899999998, 4278.7337299999999, 4278.7407700000003, 4278.7488199999998, 4278.7586199999996, 4278.7705699999997, 4278.7825599999996, 4278.79259, 4278.8002399999996, 4278.8098, 4278.8203299999996, 4278.8247000000001, 4278.8328199999996, 4278.8436799999999, 4278.8563700000004, 4278.8657000000003, 4278.8771100000004, 4278.8879900000002, 4278.8969999999999, 4278.9125599999998, 4278.9224400000003, 4278.9297399999996, 4278.9374799999996, 4278.9458699999996, 4278.9558999999999, 4278.9677000000001, 4278.9782500000001, 4278.9891600000001, 4278.9983599999996, 4279.0078800000001, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, 4279.05818, 4279.0842400000001, 4279.1054700000004, 4279.1269400000001, 4279.1490599999997, 4279.1737499999999, 4279.1917999999996, 4279.1986900000002, 4279.2098100000003, 4279.2206100000003, 4279.2332999999999, 4279.2441500000004, 4279.2527, 4279.2607699999999, 4279.2696500000002, 4279.28089, 4279.2927, 4279.3469599999999, 4279.3998799999999, 4279.4663200000005, 4279.4920899999997, 4279.5146100000002, 4279.5361599999997, 4279.5591000000004, 4279.5814200000004, 4279.6039700000001, 4279.6254200000003,
...}


I have just fixed these issues and do some more testing on the real data, so far so good.

My Energy Consumption

Follow to FigureEnergy study, I have successfully installed AlertMe at my house for 5 days now. It took a while to fix some coding issues to trigger the data to transfer from AlertMe to FigureEnergy server. Then, having observed my energy consumption a bit, I realised the data seems to be ridiculous.

Let me show you the energy figures I have observed:

Begin with the energy data this week, obtained from the AlerMe website. It showed that the maximum energy I used was approximately 18.2 kWh on last Wednesday 09 Feb 2011.

The energy consumed on the specific dates are:

As I can see from the graph, the energy consumption is sometimes suddenly rocket in a small specific period. 

I am not sure whether these energy data look right or not. It could be wrong if I install the Meter Reader of the AlertMe to the wrong cable.

On the other hand, what I have seen my energy consumption, from 09 Feb to 11 Feb , in FigureEnergy logger was unbelievable. There data are:

FigureEnergy logger displayed the energy consumption from 09 Feb 2011 to 11 Feb 2011

FigureEnergy logger displayed the energy consumption on 09 Feb 2011

FigureEnergy logger displayed the energy consumption on 10 Feb 2011

FigureEnergy logger displayed the energy consumption on 11 Feb 2011

In conclusion, there must have something wrong here, and I need to find out what they are.

The Design of Eco-Feedback Technology (Jon Froehlich et al.) - 2010

1 – What are the two most interesting things you learned from this paper?

-      It offers a wide range of background research on key motivation techniques that environmental psychology and the behavioural sciences at large have used to change behaviour such as information, goal-setting, comparison, commitment, incentive/disincentives and rewards/penalties, and feedback. The example domain is energy consumption.

-      Provide feedback on individual or group behaviours with deeply analysed the environmental psychology. The prototypes are built after deeply studying the particular behaviour as a specific behaviour has its own set of contexts and constraints which impact behaviour change.

2 – How do you think that relates to our work on FigureEnergy?

Figure Energy is design to motivate people ease of managing and learning their own energy consumption with the ultimate goal is to help them reduce their energy usages and costs. We need to apply some key motivation techniques such as rewards/penalties, commitment, goal-setting and comparison, or even combine these techniques to our system design to achieve the goal.

Intelligent agents can learn the human behaviours to give feedback, given the set of contexts and constraints which impact behaviour change. In addition, agents then can routinely learn the connection between the amount of resources users use and the consuming behaviour.

I believe these motivation techniques are the main keys to actually make the Multi-Agent Interaction work in the future.

Hypothesis Testing

1 – What is statistical hypothesis testing?

 A statistical hypothesis testing is a method of making decisions using data, directly tested through an empirical investigation. The hypothesis testing lays both the foundation of experiments and the basis of statistical significance testing.

Typically, the hypotheses of an experiment involve examining a relationship between the dependent variables and the independent variables. They specify the dependent variables to be observed and the independent variables to be controlled. An experiment normally has a null hypothesis and an alternative hypothesis. A null hypothesis (Ho) states that there is no difference between experimental treatments, while the alternative hypothesis (H1) is a statement that is mutually exclusive with the null hypothesis. The goal of the experiment is to test the null hypothesis against the alternative hypothesis and decide which one should be accepted and which one should be rejected, based on the results of any significance test. [1,2]

Significance testing technically is a process to determine the probability that the null hypothesis is true, where a null hypothesis contrasted with the alternative hypothesis. All significance tests are subject to two types of error. Type I errors refer to the situation when the null hypothesis is mistakenly rejected when it is actually true. Type II errors refer to the situation of not rejecting the null hypothesis when it is actually false. Type I errors generally are worse than Type II errors, therefore the alpha threshold to determine the probability of making Type I errors should be kept low, widely accepted at 0.05.

2 – What implications does it have on the design of our study?

We have several testing on our system to examine how effective our system design offer to users an ease of use in managing their energy consumption and saving their own usages. We can use the hypothesis testing on experimental research to make a decision on our study.

We might consider taking two experimental studies:

1 – With the use of activity annotation, users allow to tag a segment of usage from their previous energy consumption profile with specific activities. We will test to figure out whether or not using the “flat reward” (a fix amount of reward) better than using the “linear reward” (the more energy usages which users saved, the more rewards users offer)

2 – Without the user of activity annotation, We will test to figure out whether or not using the “flat reward” (a fix amount of reward) better than using the “linear reward” (the more energy usages which users saved, the more rewards users offer).

References

[1] Lazar J., Feng J. H., and Hochheiser H., “Research Methods in Human-Computer Interaction”

[2] Interaction Design – beyond human-computer interaction (corebook)

Tueday 08/02/2011

I have recently done some reading, here is a few updates:

-           Have read the paper of “TheDesign of Eco-Feedback Technology” and going to read some related reference papers.

-         Statistical hypothesis testing: have understood after reading Wikipedia and the HCI book, gonna read more on the “Research methods” book.

I now try to develop the main concept/model of the 9-month report as I think it is the focus point. Therefore I always try to get some related points from the papers/books to develop my thinking. In addition, I think it is better to start developing my thinking now (with help from supervisors) and I hope the model will be built soon.  Here are my initial development points, if you would find anything to add, please do.

Initial Concept/Model Development:

·         Each house has an agent, acted on behalf of the house owner

·         Each agent can see the daily energy consumption

·         The electricity costs vary depending on time:

·         Costs may be different from day and night

·         Costs may be different from the daily hours

·         Costs may be different based on the tariff given by the electricity suppliers

·         Users (Humans) have a list of activities (events) daily

·         Each event normally costs a certain amount of energy usage

·         Some events (human’s habits) seem to repeat in a given period (e.g. cooking, having showers). (This is one of the feature from Figure Energy)

·         Users want to reduce/minimise costs

·         In an organisation with the given number of agents, agents can negotiate, form with others to maximise the social welfare

·         The optimal social welfare is to minimise the electricity costs

·         Constraints (for coalition formation)

·         Human interference (e.g. users might need the specific activity at the specific time)

I will read and discuss more for further development.

Thusday 03/02/2011

Having tested yesterday, the "Zero-gap" seems to disappear now, great!

Today is time for reading. I am now reading on the Interaction Design core book to understand more about Hypothesis testing.

to be updated ...

Wednesday 02/02/2011

- Working on "Zero-gap" now but I still have no idea how to fix it :(

- I finally fixed this issue.

Tuesday 01/02/2011

Having had a meeting this morning, here is the priority list for the next few days:
1 - Deploy and Commit (using Fabric)
2 - Fix "Zero-gap"
3 - maybe add more icons (if find any)
4 - Install AlertMe

Some reading concerned in parallel:
- Textbook:
      +  Reading about "Hypothesis testing" (in the main HCI core book)
- Paper:
      + "The Design of Eco-Feedback Technology" and the references.

I also need to start to write the 9 month literature review, begin with do some summary writing about papers or textbook I would have read.

I also had got through the FigureEnery overview and understand more about the Planner and Analyser widgets.

To be updated ...