Indicación del rendimiento para el inversor

ABM has been used to simulate the effect of external factors on the behaviour and energy consumption of occupants. These external factors may include energy management policies, technologies and interventions. For example, Zhang et al. [62] compare the effectiveness of energy solutions, including an automated lighting strategy and a human-controlled one. Although this model includes an energy awareness attribute to simulate realistic lighting behaviour, the energy solution they propose and test is not aimed at enhanc- ing the energy awareness of the occupants, but only testing the automated strategy. Therefore, it may not be categorised as an energy intervention. An energy intervention approach is proposed in [64], where the research aims to test a number of building management and control approaches. One of the tested approaches includes a proactive meeting relocation capability. It suggests changing meeting rooms to smaller rooms or rooms that were pre- viously occupied (i.e. previously heated) to save energy consumption in a university building. The occupant agents may or may not accept the reloca- tion suggestion based on the meeting constraints and their energy conscious- ness. However, the model does not capture the change of occupant energy consciousness/behaviour in effect of the proactive approach after several in- cidents of relocation. This contradicts the aim of energy interventions, which is changing occupant actions and decisions to reduce their consumption.

Zhang et al. [22] simulate the learning experience of household agents as a result of smart meters usage in the United Kingdom. For this purpose, they use the behavioural learning theory where households learn through repe- tition and conditioning. The formula is used to determine when the agents change from inexperienced to experienced smart meter users. In this model, energy consumer agents are modelled as a whole household that owns and

uses the smart meter. This type of modelling not only affects the realistic occupancy and behaviour simulation as mentioned in Section2.2.1, but also causes the loss of individual level dynamics where the intervention may af- fect household members in different ways.

The ABMs in [13], [58] study the effect of discrete interventions such as energy training and workshops along with peer pressure effect, which helps the diffusion of the green behaviour in office buildings. As mentioned in Section2.2.1, these models do not generate detailed occupancy and activities of occupants. Therefore, the only way to simulate the discrete intervention is by randomly selecting the affected individuals, and changing the energy consumption level based on the assumed success percentage of the interven- tion. Besides, it is not possible to simulate accurately the effect of continuous interventions such as feedback or messaging interventions.

One of the models that simulate continuous interventions is proposed by Anderson and Lee [14], who compare the effect of individual and compar- ative – to neighbours for example – feedback. The model stochastically de- termines the possibility of checking the feedback, which causes change in occupants’ behaviour. Feedback interventions are also studied in Jensen et al. [27], where the intervention effect is modelled using an asymptotic equa- tion. The behaviour level is changed toward an incentivised target with a specific rate, which represents the intensity of the intervention. Similarly, Lin et al. [24] and Lin et al. [63] evaluate the effectiveness of tiered pricing in office buildings. The change of the energy awareness of occupants depends on the level of change in the price. Although these models [14], [24], [27], [63] sim- ulate continuous interventions at occupant level, the used behaviour change equations assume the same effect of the intervention in all cases. However, the effectiveness of interventions may vary based on how often the occupant uses the intervention, whether he/she is interested in it and his/her social and psychological characteristics. The ABM proposed in this thesis simu- lates realistic interaction of occupants with the energy intervention based on their social characteristics and interest in it, and changes the awareness of the occupants based on how often they complied to it.

2.4.2

Energy Feedback Systems

As mentioned previously, feedback is one of the interventions that aims to help occupants save energy. Consuming energy is considered abstract and invisible as it is used indirectly to perform daily tasks [92]. Therefore, it

2.4. Energy Interventions 37 is agreed that giving people information about the amount they are using makes them aware of their consumption and ultimately allows them to con- trol it. Direct feedback is available in various forms, including meter read- ing, direct and interactive feedback via monitors, pay-as-you go meters and plug/ appliance meters [91]. However, with the advancements in sensor and communication technologies, direct and interactive feedback is now the most common [29]. For example, in response to the European Commission plan to reduce 20% of the Union’s energy consumption [93], the United Kingdom has installed 8.5 million smart meters (along with feedback displays) up to 2017 [94].

Energy Feedback Systems (EFS) have been widely researched to study their effectiveness and users’ interaction with them. For example, the ef- fectiveness of simple energy displays (stationary and portable) was investi- gated in [30]. The study shows that energy displays resulted in an average of 11% energy reduction and increased the energy awareness of occupants. Besides, commercial feedback systems were assessed qualitatively in [95] by asking people about the motivation of owning display systems, ways of us- age, observed behaviour change and limitations of usage. Along the same lines, Karjalainen [96] systematically reviewed the different ways of present- ing feedback. Several user interface prototypes were developed with varied comparison types, units of display, disaggregation levels, presentation types and time scales. They found that presentation of energy costs, appliance con- sumption and historical comparison are the most preferred by users.

Although these studies showed that EFS play a role in increasing occu- pant awareness, many studies highlighted a number of limitations. For ex- ample, Strengers [31] observed that a considerable number of users struggled in understanding the displayed data and converting them to meaningful in- formation. This is because the displayed data are absolute and not related to the surrounding context. The same conclusion was reported in [97] where people wanted more context, such as occupancy and temperature to inter- pret high/low consumption levels. In response to this challenge, a num- ber of studies suggest to relate energy consumption to daily activities either by annotating consumption graphs with activities [33], or using calendars as an artefact to help understand consumption [98]. Similarly, Castelli et al. [32] propose to use the location of appliances and occupants, which they call ‘room context’. This helps identify energy wastage, match consumption with occupant presence and link consumption with everyday activities.

view users as micro-resource managers [31], [99] who are expected to anal- yse the displayed data and change their behaviour– such that it meets their preferences, everyday needs and financial and environmental goals. Based on this, Pullinger et al. [99] identify one more specification for EFS, which is explaining what the information means in terms of behaviour change. In addition to detailed energy consumption data, this service requires collect- ing environmental data and Artificial Intelligence (AI) analysis techniques, which are not provided by existing EFS. In this thesis, we try to fill-in this gap by proposing the idea of an energy messaging intervention, which pro- vides occupants with sensible messages that tell them what to do to reduce their consumption, instead of only giving them the amount of energy they are using. We identify the technologies and techniques available to collect and analyse the required data, and test the effectiveness of this approach in an energy simulation model in Sections5.1.4and5.2of Chapter5.