Based on research literature and our own surveys, energy consumed by household domestic appliances can be divided into six portions: cooking, home computing, consumer electronics (e.g. TVs), wet appliances (e.g. washing machines), cold appliances (e.g. refrigerators) and lighting. Except for home computing, the other five parts share an approximately equal portion of domestic energy consumption, with consumer electronics taking up a slightly larger portion (12).

The algorithm in the REACT system will examine the household’s usage patterns of each individual appliance and provide system-generated suggestions regarding their electricity usage pattern to achieve reductions in electricity bills where possible. The information gathered will include specific time of habitual usage of each household electrical appliance, the duration of usage, and the characteristic category of that particular appliance as depicted in Fig 2.

2.2.1 Categorizing Household Appliances

Other than by type, appliances can also be grouped by their flexibility of use with respect to time. Based on the feedback from our survey, it is reasonable to divide household appliances into three categories - Priority, Flexible and Automatic.

Fig 2: Household appliances by time-flexibility of use

2.2.1.1 Priority Appliances

Priority appliances refer to devices whose usage depends almost completely on user discretion and are unlikely to be flexible, such as electric hobs and consumer electronics. As these appliances are generally switched on due to the customer’s demand and are fixed in time, the algorithm will not make energy-saving suggestions. However, these appliances will still be monitored to provide power usage information to inform the user of the power usage in this category of appliance.

2.2.1.2 Flexible Appliances

Appliances in the flexible category are those which are generally regular and can be deferred to run at different times of the day, such as electric tumble dryers, washing machines, and dish washers. Consumers generally are not very particular about when these devices execute as long as the work can be completed before a consumer-specified deadline. By receiving real-time prices, the algorithm can choose the best time to execute these appliances based on historical prices.

2.2.1.3 Automatic Appliances

Automatic appliances are those which require no human input and can be run autonomously by the REACT system. These include refrigerators and electrical heaters, which, once set, require no further human control.

2.2.2 Algorithm Governing the Control of Appliances

Fig 3: Flow of information for the real-time pricing system

Given the information on consumer demand for electricity and the producer’s capacity to supply, it is possible to determine a price which will maximise benefit for both consumers and producers. This was shown to be attainable by researchers from The University of British Columbia, where a smart power infrastructure was designed which allowed two-way communication between the suppliers of electricity and households (13). A utility maximisation algorithm for their system allowed suppliers to determine the real-time price for the grid. The price was calculated according to the current electricity price and household expected load, which were updated on an hourly basis.

As the REACT system is targeted at households, the determination of real-time price by the supplier is out of the scope of our project. However, the algorithm will discuss how, given the historical and real-time prices of electricity, the algorithm can predict the time when electricity price is at its lowest, and therefore make decisions for devices in the Flexible and Automatic categories.

The algorithm in the REACT system will primarily attempt to predict future prices by observing historical prices. In Fig 4, it can be seen that the price for electricity is lowest around 3am. Therefore, assuming the user has included this period as part of the target schedule for Flexible appliances to execute, the algorithm will aim to switch on the appliances at 3am, thereby saving money for the user. However, taking into account that the REACT system is intended for widespread use by consumers, a completely deterministic approach based on past prices may be problematic. If all users predict that electricity prices are cheapest at 3am, and all the appliances are made to run in that period, this will simply result in a demand peak at 3am (and correspondingly higher prices). This defeats the purpose of the REACT system in the elimination of peak demand.

The solution to this problem would be to use the historical price chart to create a probability density function (PDF). The PDF would have a maximum corresponding to the lowest historical price on the previous day, and a zero corresponding to the maximum previous-day historical price. This can be done by calculating the PDF as:

This is shown in Fig 5 above.

With the PDF, the REACT system would be able to carry out a random trial to determine when to execute its appliances. As the PDF puts heavier weightage during the times when energy is cheapest, the system will tend to select these periods more frequently. However, it will balance these cheap periods by occasionally selecting the more expensive periods to run the appliances, so that not all households will behave in an identical manner. The net effect is that while some households will execute their Flexible appliances during the more expensive periods, most households will be running their appliances during the cheaper period and therefore realise significant cost savings.

When looking at the PDF, it is also important to remember that not all appliances will be controlled by the REACT system. So while it seems that no appliances would be running at 5pm, in fact, usage at 5pm would likely still continue to be a peak, simply because consumer demand at that period would be inflexible and probably part of the Priority category.

With the price information from the suppliers, the pseudo code of algorithm is as follows:

1: initialization;

2: label loop forever;

3:check category;

4:if (the appliance is under priority category)

5:do nothing;

6:else if (the appliance is under flexible category or under automatic category)

7:check duration of appliance;

8:check user-specified execution timing range;

9:fetch historical prices from memory or smart grid, within user-specified range

10:calculate pdf using historical prices;

11:conduct a random trial using pdf to determine appliance execution timing;

12:store appliance execution time, for triggering later

13:send update to electricity supplier detailing scheduled appliance timing;

13: go to loop forever;

With the algorithm above, the REACT system is able to find times when electricity prices are low and also manage to avoid the creation of new demand peaks through the use of probability to determine the specific execution timing of appliances.

2.2.3 Monthly Report Algorithm

In addition to the real-time control of appliances, the monitoring of consumer power usage is another major part of our system and provides comprehensive feedback to consumers in the form of a monthly report. The report aims to provide information such as each individual appliance’s current category (Automatic, Flexible or Priority), the household’s daily power consumption data, and the comparisons between the consumer’s habitual power consumption versus the power consumption savings if the user had agreed to the system’s energy-saving suggestions. The aim of monthly report is to present information to the consumer in a clear manner in order for the consumer to best understand and control their energy usage.

2.2.3.1 Input to Monthly Report

The monthly report algorithm records the power consumption from each appliance through the sensing modules as well as the real-time pricing data from the electricity supplier. To achieve a balance between memory usage and precision, the usage and pricing data will be stored on a minute basis.

2.2.3.1 Output of Monthly Report

With the power consumption of each appliance and pricing data recorded, the monthly report algorithm has all the data it needs to calculate the daily and monthly power consumption of the customer, both in terms of raw usage and cost. The REACT system can then give energy-saving suggestions and show how acting on these suggestions could save the consumer money. This is useful as consumers are generally unaware of the impact of altering their energy usage habits on the price they pay.