Evaluation in terms of energy usage

For HAS, the system effectiveness is reflected by the amount of energy that can be reduced with active smart strategies. Hence, we select fixed schedule smart plans as specified in section 4.1as the comparative basis. The point of this phase is to verify the role of user context in improving efficiency. Power or energy consumption is primarily affected by the period devices are in use. By shutting down the appliances while not in use, we suf-ficiently save to-be-wasted energy. In this sense, increasing the time power-consumed

devices in off or standby mode will reduce the energy usage at a certain amount. Experi-ments and observation are made based on specifications of German use case. We presents results in two main groups of appliances: heating/cooling system and lights.

(a)Heating/cooling system is turned OFF

at 8AM daily. (b)Heater is turned ON at 4:30 PM daily.

Figure 17: History graph of heating/cooling system status on a fixed schedule scenario.

Switch on/off heating or cooling system by controlling the thermostat

Consider thermostat controlling, as shown in Figure 17, in fixed schedule scenario, the heating/cooling is turn on/off at a specific point of time during the day, which is 16:30 (4:30 PM) and 8:00 AM respectively. For out of ordinary circumstances, the system effi-ciency can benefit from abnormal reaction if events happen out of this normal functioning range, for example, if the occupant has an event scheduled at, for example, 7 AM, which indicates that the occupant will leave house at an actual time around 7 AM, one hour earlier. In this case, the amount of energy user can save is calculated as follow:

P_saved= (8−Actual Leaving Time)∗(Heating/Cooling Energy per hour)

(Actual Leaving Time<8) (5.1) Similar logic applied to the case where user may get home later than they use to, which leads to:

P_saved= (Actual Arrival Time−16.30)∗(Heating/Cooling Energy per hour) (Actual Leaving Time>17) (5.2) The study only considers cases where ETA to home is after 17:00 due to the specification that heating or cooling system should be turned on at least 30 minutes before arrival for

user comfort. An interesting notice from this evaluation is that the energy that can be reduced is proportional to the difference between abnormal behaviors and normal behav-iors. In another word, the calculations indicate that the more difference the user unusual behaviors, the better adaption the system can reach.

Follow the same principles, we conduct experiments in within a limit of one week to ob-serve the reaction of system, where user occurs to have events scheduled at different point of time that can affect the schedule of heating or cooling system. The event specifica-tion of test cases are shown in Table 8. According to this, we present observed results inFigure 18, Figure 19, andFigure 20corresponding to three test cases described in the following table.

Table 8:Events extracted from user’s calendar.

Event Date Start time End time ETA to home (mins) E1 2019-06-02 07:00:00 09:00:00

-E2 2019-06-03 18:00:00 20:00:00 30 E3 2019-06-04 13:00:00 14:30:00

-Figure 18shows the on/off period of the heating/cooling system according to fixed sched-ule from date01.06.2019to05.06.2019. Every day the heating/cooling system is turned off at 8:00 in the morning when the occupant leaves the house and turn on at 14:30, 30 minutes before the time that the occupant is supposed to be at home.

InFigure 19, we observe this process with user context integrated plan enable. In the case of E1 and E2, the time that user leaves the house and arrives at home affects the operation mode of the heating/cooling system. On the day that user leaves at 7:00, system is turned off at that point of time, instead of 8:00 as scheduled and stays off until 16:30.

And we have test case with E3, although user has event scheduled, the event does not af-fect the heating/cooling system because it actually ends before the heating/cooling system needs to be switched ON, which makes perfect sense in this case. Finally, we have the results as shown inFigure 20. The periods where appliances stays OFF longer reflects an improvement in energy consumption, in another word, we observe an improvement in the energy efficiency of the targeted smart home system.

Following the same principles, we achieve similar results in the case of controlling lights with and without considering user context. The difference in working periods of devices

Figure 18:On/off period of heating/cooling system - fixed schedule.

Figure 19:On/off period of heating/cooling system - context adapted.

depends on the circumstances and is not fixed, in summary, can be calculated as:

P_saved= [(8−Actual Leaving Time)∗(Energy per hour)

+ (Actual Arrival Time−16.30)∗(Energy per hour)]∗number_o f_devices (5.3) Using parameters described inTable 8, we applied equation 5.3 to calculate the amount of energy can be saved in this specific scenario. In this case, the average energy consumed per hour for heating/cooling system is about 1.5 kW/h (Oasis Energy, n.d.). For German household, we assume user will have the heating system on from 16:30 until 8:00 next morning, which is equivalent to 15.5 hours/day or 15.5∗7=108.5(hrs/week)the heating

Figure 20:Compare on/off period of heating/cooling system.

system is in ON status (consuming energy). In our experiment, this ON time is:

T_heating_ON=108.5−[(8−7) + (20−16.5)] =104(hrs)

In normal circumstances, we assume only energy consumption during weekdays are dif-ferent because user stays at home during weekends, thus amount of energy that can be saved is:

P_saved=4.5∗1.5(kW/h) =6.75(kW h)

Or in another way of expression, in this scenario, by activating user-adaptive HAS, we can save about(4.5/108.5) =4% of the energy consumption per week.