Permuta Financiera de Intereses

In the previous chapter, it was shown that social parameters affect the en- ergy efficiency of the family. Although the previous model does not simulate family pressure, we showed that energy waste in large families is less than small families. On the basis of this conclusion, the current experiment tests whether the repression family-level intervention is more effective in big or small families. For this purpose, the family-level intervention is applied on (a) a two-occupant family (two adults 25-39 years old both with full-time job), and (b) a four-occupant family (two adults 25-39 years old both with full-time job, and two children). The initial threshold d of both families is 0, and the initial number of green and waster occupants is equivalent ((a) FD and (b) FFDD). Figures4.9aand4.9bshow the consumer types convergence of scenarios (a) and (b), respectively. Figure4.9cshows the resulting energy saving percentage when compared to the no-intervention scenario (Ir = 0) and the convergence time, which is the time it takes the family to reach a stable state where the occupants are no more affected by each other.

In Figure4.9c, at intervention intensities 1 and 2, the percentages of sav- ing for big families are 9% and 16% respectively, which are more than that of small families (i.e. 1% and 11%). This is also observed in the awareness type convergence (Figures4.9aand4.9b) where the ‘4 green occupants’ category is

4.3. Experiments and Results 89

2 Green Occupants 1 Green Occupant No Green Occupants

0 1 2 3 4 I_r 0 20 40 60 80 100 Percentage of Households

(a) 2 Occupant Family (FD)

(b) 4 Occupants Family (FFDD)

4 Green Occupants 3 Green Occupants 2 Green Occupants 1 Green Occupant No Green Occupants

0 1 2 3 4 I r 0 20 40 60 80 100 Percentage of Households (b) 4 Occupants Family (FFDD)

FIGURE 4.9: Effect of Family-level Intervention on Two- and Four-Occupant Families – d = 0 (Continued in the next page)

more dominant in (a) than the ‘2 green occupants’ category in (b). However, at intensities 3 and 4, the savings of small families are 21% and 25% respec- tively, which dominates that of big families (i.e. 16% and 15%)(Figure4.9c). Besides, all of the occupants in scenarios (a) and (b) converged to green occu- pants as shown in Figures4.9aand4.9b. This is explained by the lower con- vergence time of small families (Figure4.9c). This means that a higher inten- sity intervention converges small families quicker than big families, which consequently leads to higher saving. Besides, low intensity interventions can lead to maximum saving in big families.

4 Occupants Saving 2 Occupants Saving 4 Occupants Convergence Time 2 Occupants Convergence Time 1 2 3 4 I_r 0 10 20 30 40 50 Percentage of Saving 0 8 16 24 32 40 48

Convergence Time (week)

(c) Saving and Convergence Time

FIGURE 4.9: Effect of Family-level Intervention on Two- and Four-Occupant Families – d = 0

4.4

Discussion and Insights

The model proposed in this chapter simulates peer pressure effect on energy awareness levels and consumption of families. The peer effect behaviour of occupants is underpinned by the well-established human behaviour theo- ries by Festinger[38]–[40] opposed to other models that do not use existing theories [58]. Besides, the existing models [7], [14], [26], [27], which are de- veloped for residential and commercial communities vary the structure and type of peer networks, which is not applicable for families. The developed peer pressure model uses behaviour theories that are adapted to comply with family pressure effect on energy consumption in households. In addition, as the model is built upon the daily behaviour model, it separates the change in energy use behaviour attribute from the daily activities of occupants.

The experiments presented in this chapter showed how the model can be used to analyse the effect of interventions in a number of significant scenar- ios. We proved that the promotion family-level intervention is more effective than the repression intervention in a family where the waster occupants dom- inate. This implies that it is better to give incentives and rewards for mem- bers of waster families to encourage them to adopt the green behaviour. For the occupant-level intervention, we showed that even when green occupants dominate in a family, one intervention per year is not enough to prevent the waster occupant from affecting the green occupants. This indicates that occu- pants need to be continuously targeted by interventions even if the number

4.5. Summary 91 of green occupants is high in the family.

In the last experiment, we showed that the family-level intervention can result in maximum saving at low intensity in big families as opposed to small families. While a high intensity intervention is more effective in small fami- lies as it leads to a larger and quicker saving than big families. This indicates that more effort is needed when targeting small families, while big families can reach their maximum saving with lower effort. These implications are compatible with the ones obtained from the experiment that studied family size in Section 3.4.2 of Chapter 3 and discussed in Section 3.5. The recom- mendation in Chapter3was to target small families with focused and high intensity interventions, and big families with low intensity and distributed interventions. This is because small families waste more energy as proved in in Chapter3, and save more when targeted with high intensity interventions as shown in this chapter, and big families waste less and save the maximum when targeted with low intensity interventions.

The developed model can be used to repeat these experiments with var- ied social parameters, thresholds and intervention types to obtain the most effective intervention in every case.

4.5

Summary

This chapter has presented a model that simulates the effect of peer pres- sure on the energy consumption of households. Behaviour change due to peer pressure is modelled using Festinger theories (informal soaicl commu- nication theory, social comparison theory and cognitive dissonance theory), which are compatible to a family setting and lead to an uncomplicated model. Besides, these theories are formalised by adapting Granovetter’s Threshold Model. The occupant agents change their energy behaviour every period of time based on the observed behaviour of other occupant agents around it. The model also includes an abstraction of two types of energy interventions (occupant-level, and family-level interventions). This is done to study the effect of peer pressure on the results of interventions. The presented exper- iments prove the conceptual validity of the model – such that it reflects the used theories. In addition, they show how the model offers an analytical tool for governing bodies to analyse the effect of interventions and decide how to target different families to get the best desired results.

The next chapter presents an energy messaging intervention, which is cat- egorised as an occupant-level intervention. The intervention will be tested

using a third model that is built over the daily behaviour model presented in the previous chapter and the peer pressure model presented in this chapter.

93

Chapter 5

The Messaging Intervention Model

After presenting the peer pressure model that simulate realistic occupant-to- occupant interaction, this chapter models the occupant-intervention interac- tion. The chapter also proposes a messaging intervention that combines the technologies used for automated control and the service of providing energy feedback. The proposed intervention is implemented in a third layer of the proposed complete Agent-Based Model (ABM) (Figure5.1). The messaging intervention model is built upon the daily behaviour model and the peer pressure model.

FIGURE5.1: The Messaging Intervention Model from the Lay- ered ABM

In this chapter, we show the strengths of designing the complete model in a layered onion-like structure for the following reasons:

• The messaging intervention model takes advantage of the detailed daily behaviour and energy consumption data generated by the core model, and the realistic family pressure simulated in the peer pressure model. The previously developed models enable the simulation and detection of energy waste incidents, which are used to test the messaging inter- vention. Besides, they are necessary to simulate realistic interaction of occupants with the intervention.

• The layered onion-like structure enables plugging and unplugging var- ious intervention types and mechanisms taking advantage of the data generated by the models. This makes the complete model a useful tool for policy and decision makers to design customised energy interven- tions.

The following section presents the proposed intervention, then, the model formalisation and design are presented. Next, a number of experiments that assess the intervention are presented and discussed. Findings reported in this chapter are published in [130].

5.1

The Proposed Messaging Intervention

Based on the literature reviewed in Section2.4, we present a messaging in- tervention that is a middle solution between Energy Feedback Systems (EFS) and Energy Management Systems (EMS). Therefore, instead of providing the amount of energy being consumed or comparing the household consump- tion with similar ones as in EFS, the intervention provides the occupant with real-time messages about their current energy wastage and recommends ac- tions to reduce their consumption. This is done by relating the energy con- sumption of appliances with the context of the house, including occupant presence, activities and schedule, as well as environmental data. This en- ables the detection of energy waste incidents in which the intervention can recommend to reduce this waste based on the occupant state. The approach in this intervention is to avoid taking automatic actions opposed to the gen- eral approach in EMS. This is done to maintain the occupants’ comfort allow- ing them to take decisions whether to comply with the messages or not. An example of real-time messages would be: "Your television in the master bedroom is now ON while nobody is there, it is recommended that you turn OFF devices while not in use", or "The lights in the living room are now ON while there is enough day- light in the room, you can take advantage of natural daylight to reduce your energy consumption".

The following sections (1) detail the type of appliances that was imple- mented in the simulation model, (2) define a messages pushing strategy/ heuristic to control the rate and number of messages to be sent to occupants, (3) present the factors that affect occupants energy consumption behaviour, including compliance to the waste messages and (4) present different en- abling technologies and techniques that may be used to obtain and forward the messages in reality.

5.1. The Proposed Messaging Intervention 95