lnvestigating Energy Prosumer Behaviour in Crowd Energy Using an lnteractive Mod

Christian Ritzel | Anton Sentic＊

Abstract—Smart living labs such as the one located in Fribourg (Switzerland) focus on improving wellbeing and furthering knowledge related to building the district of the future on a technical and social level.Therefore, smart living labs represent an experimental platform/space where sustainable production and consumption strategies can be tested in a protected environment.A significant change in the socioeconomic production and consumption sphere can be expected by the rise of the so-called energy prosumer.Accordingly, this article presents an interactive model for the experimental investigation of energy prosumer behavior.In this context, two potential experiments on investment and trade decisions are briefly outlined.Since (behavioral and economic) experiments are usually conducted under controlled conditions in experimental labs involving mainly undergraduate students, the presented interactive model is flexible and mobile, providing the advantage to conduct experiments nearly everywhere involving everyday citizens.

1. Background: lnfrastructural and Organisational Aspects of the Crowd Energy Concept

One possible solution to fulfil the aims of the Swiss Energy Strategy 2050 and to increase (energy) selfsufficiency and resilience is to promote the rise of the so-called energy prosumer and energy prosumer communities.A prosumer can be considered as an agent who produces, stores, and consumes its own energy, for instance through a photovoltaic (PV) system on its own rooftop[1].On a global level, (energy) prosumers can play an important role in future development pathways for energy systems, particularly in the case of energy network decentralization and heterogeneity[2].While this scenario is only one of multiple possible ones, and the ultimate direction of energy system development is far from certain, prosumer communities offer a potentially efficient and intriguing approach to solving energy justice challenges[3],[4]as well as contribute to balancing the “energy trilemma”[5].The “energy trilemma” concept is defined by [6] as a set of law and policy challenges stemming from the need to balance competing demands of politics (energy security), economics (energy finance), and the environment (need for climate change mitigation).Prosumers and prosumer communities exhibit high potential to create positive externalities for society and environment (e.g., local job creation, increasing community solidarity/identity, avoidance of potential land use because PV systems are constructed on already existing rooftops)[7].

Consequently, (individual) prosumers and prosumer communities are an important stakeholder of future Swiss smart grids with a vital role in peak demand management[8].In this context, a fundamental example for the crowd energy concept involving the energy prosumer is provided by [1], which is “Crowd energy is the collective effort of individuals or profit or non-profit organizations, or both, pooling their resources through online information and communications technology-applications (ICT-applications) to help to implement the energy turnaround.This implies both, the concept of decentralization (production, storage, and consumption of renewable electricity) and a substantial change in society, economy, and politics.”

The crowd energy concept leads to a change in the role of the energy consumer from pure consumption to prosumption (generation, storage, and consumption/load of energy).Fig.1 presents the heart of the crowd energy concept—the prosumer-based intelligent generation-storage-loading (iGSL) cell.

Technically speaking, the introduced prosumerbased iGSL cells are the basic elements of the crowd energy concept.Pooling numerous cells in a network,connecting them by communications lines, and finally processing the information in a crowd management system yields to an iGSL cell crowd (see Fig.1).In this context, it should be emphasized that the structure of the crowd does not have to be fixed.It is rather the case that the iGSL cells can and should vary over time, which refers to the term “Swarm-Electrification”.In other words, the prosumer-based energy crowd is able to grow dynamically[1].Basically, the crowd energy concept can be realized in a virtual network where iGSL cells are i.e.distributed over Switzerland, or in close proximity, which refers to an energy neighbourhood.In this context, [7] provides an overview of structural attributes of prosumer-based energy crowds as presented in Fig.2,where dots represent iGSL cells (energy prosumers), lines represent a transaction of prosuming services, and circles represent organised energy crowds (prosumer communities).Fig.2 (a) represents a peer-to-peer model, in which individual iGSL cells are interconnected directly with each other for the purpose of buying and selling energy services.Figs.2 (b) and (c) represent structured models involving iGSL cells connected to microgrids.Here, iGSL cells provide energy services to a microgrid that is in turn connected to a larger grid.Fig.2 (d) represents an iGSL cell (prosumer) group model, in which a group of iGSL cells pools their resources or forms a virtual network (a virtual power plant).Due to their organisational form and the utilised technological artefacts, energy crowds can easily be imagined as smart grids and/or decentralised energy networks[2], depending on the degree of crossborder crowd integration and the structural and institutional overlaps with the “conventional” energy grid.In Fig.2, Figs.2 (a) and (c) can be considered examples for decentralised energy networks, while Figs.2 (b) and (d)show a higher degree of cross-border integration.

Fig.1.Intelligent generation-storage-loading (iGSL) cell[9].

Fig.2.Structural attributes of prosumer communities[7].

However, despite the demonstrable promise and benefits of energy crowds, widespread diffusion of the concept requires solving multiple challenges: Questions of technology utilisation, IT management and IT security,governance, business model development, and user behaviour.In this article, the authors focus on the latter by presenting an approach to utilise behavioural and economic experiments in order to model the behaviour of individuals within an energy crowd using an interactive crowd energy model/diorama.Accordingly, the remainder of this article is organised as follows: Section 2 deals with the topic of utilising technology and economic experiments for educational purposes.In Section 3, first, the old non-interactive physical crowd energy model/diorama is briefly presented (subsection 3.1).Second, the new interactive crowd energy model/diorama is presented and discussed in detail (subsection 3.2).Finally, Section 4 provides a summary of the research outlook and possible further steps.

2. Utilising Technology and Economic Experiments for Educational Purposes

One possible way to make economic laws tangible is to use experiments[10].Like other scientists, economists observe events in nature, devise theories to explain their observations, and evaluate their theories in the light of additional evidence[11].However, even until the end of the twentieth century many scientists emphasised that it is not possible to use experiments in economics.For instance, [12] stated that “Unfortunately, we can seldom test particular predictions in the social sciences by experiments explicitly designed to eliminate what are judged to be the most important disturbing influences.Generally, we must rely on evidence cast up by the ‘experiments’that happen occur.” Likewise, [13] emphasised that “Economists cannot make use of controlled experiments to settle their differences: They have to appeal to historical evidence.”

In a controlled experiment (e.g.in biology), the random allocation of subjects to treatment and control groups eliminates the so-called selection bias.Therefore, when treatment and control groups differ significantly after the treatment, no other factors than the treatment itself should cause this difference[14].Even pioneers such as [15]were initially sceptical that economists are able to isolate a causal effect because “[…] it is known that in his choices of methods, the economist can hardly use laboratory techniques like in natural sciences.On one hand, data from the real world is necessarily the product of many influences other than the ones he is trying to isolate.On the other hand, non-intentional variables cannot be held constantly or eliminated in an economics laboratory because the real world of human beings, firms, markets, and governments cannot be reproduced artificially and controlled.” Nevertheless, [15]’s experiment on a double auction market can be considered as a breakthrough in analysing imperfect markets experimentally.Already at that time,Chamberlin realised the pedagogical benefits of economic experiments, making dry economic theory more tangible.Even if the discussion on the methodological justification of experimental economics will likely never end, especially the advantage making dry economic theory more tangible for students confers this research strand its eligibility.Therefore, today it is quite common using economic experiments in classrooms (e.g.see[16]).A classical (classroom) experiment is the so-called public goods game, which represents a social dilemma(social dilemmas are situations in which each of a group of independent people faces a conflict between the maximization of personal gain and collective interest[17]).In this experiment, n members of a group are endowed with E units of a private good.Independently of each other, each individual (subject) i must decide which amount of his or her initial endowment will be contributed to a public good.The individual contributions undertaken by the subjects are summed up and then multiplied by factor M ＞0.The total (group) contribution is then split evenly among group members.The final (monetary) payoff of group member i can be formalized as follows:

As the public goods game example has shown, economic experiments are a great way to introduce students to economic theory.They are fun for the students and for the teacher[16].Accordingly, we want to take account of this didactic aspect, combining economic experiments with technology.For this purpose, we experimentally model the crowd energy concept, using i) a modified version of an augmented public goods game in accordance to [22] for the investment decision and ii) a modified version of a double auction market in accordance to [23] for the energy trade decision.The physical crowd energy model (see Section 3) supports subjects’ imagination by visualising a local energy crowd.Although, economic experiments exhibit high internal validity, they are criticised when it comes to external validity, since economic experiments designed for scientific purposes are usually conducted with undergraduate students under controlled conditions in experimental labs of universities[24].A flexible and portable crowd energy model allows the authors to overcome this shortcoming,involving everyday citizens, e.g.at fairs or open house days.

Furthermore, the utilisation of such a model in order to transfer knowledge from researchers to the general public and gather the general public’s feedback would be in line with Open Innovation principles[25]by extending the range of participating actors in the innovation process.While the model itself as well as the interactive scenarios described in Section 2 would still mainly represent a one-way information flow, actors’feedback could be gathered in the form of focus groups and dialogue session[26]as well as short on-the-spot interviews with participating and/or interested actors[27].By testing ideas related to the crowd energy concept with prospective real-life users far in advance of a real-life rollout, researchers could adapt the practical outcomes of conceptual developments as well as democratise the entire innovation process.

3. Physical Construction of the lnteractive Crowd Energy Model

One potential and promising way for the presentation of an energy crowd is the use of a physical model(the crowd energy diorama), serving as a visual centrepiece, an information provider, and, for experiments, a framing device and a visual medium for communicating participants’ actions in the course of the experiments.In parallel with the framing function, the model could also serve as a focusing device for knowledge transfer[28]from researchers towards members of the public (at exhibitions, trade fairs, and similar events) as well as experiment participants.In the following section, the authors will present the already existing crowd energy model, described as the “old” model, followed by a concept for a new crowd energy model with the ability to serve as a physical centerpiece and focusing device for the experiments described in Section 2.At the end of the section, the authors will sketch out how the model could be used for the real-time visualisations of the experimental outputs.

3.1. “Old” Crowd Energy Model

The main inspiration for the proposed interactive model is a demonstrator model for crowd energy developed in order to showcase the concept’s principle—as such, the model was not really interactive, but could only illustrate a 24-hour cycle (day cycle) for an energy crowd comprised of four actors.The model itself was built using readily available materials: A diorama showing a section of a street with three residential buildings and one commercial establishment was built from model train accessories while the electronic components were off-the-shelf DIY items and the central control unit was a Raspberry Pi toolkit computer, as shown in Fig.3.

The software managing and operating the model was developed using the “C” programming language and could be run using a web browser; for exhibition purposes, this was done on an Apple iPad table computer.It is required to both start and stop the experiment,further; a number of adjustments can be made in order to change the parameters of the simulation, both related to the technical parameters (surface of PV cells, number of tenants in a particular building, etc.) and user behaviour (level of cooperation within the crowd, energy saving/storage preferences, etc.).In order to simulate various activities and processes in the model, multi-coloured LEDs and LED strips were used, as illustrated in Table 1.

Fig.3.Photograph of the first crowd energy model.

Table 1: List of visual indicators on the initial crowd energy demonstrator model (own presentation)

While this model, for the initial rollout of the crowd energy idea, represents a quite usable demonstrator,particularly when presenting crowd energy to non-specialist actors, such as visitors of an energy fair.However, it is ultimately somewhat limited by its nature as an unidirectional interface (even though some of the demonstration parameters can be changed, this needs to be done by the model operator, removing the possibility for other actors to directly engage with the model).

3.2. Concept for a New lnteractive Crowd Energy Model

Based on these considerations as well as planned behavioural experiments (see Section 2),the authors have planned to develop a second crowd energy model with extended functionalities, where the demonstration function of the first model would be extended towards a bi-directional interface, with the model indicating actors’ actions and interactions with the model in real-time.The model would consist of a physical diorama showing a neighbourhood consisting of multiple buildings, fitted with LED indicators indicating various activities and processes taking place within the simulation, a base for the model and multiple tablet computers serving as interfaces.Fig.4 shows a concept sketch of the model.

The model would consist of the following components:

·A physical diorama showing multiple buildings in a neighbourhood, representing the simulation (energy crowd) participants.The diorama will be made from model train buildings in scale H0 (1:87), supported by other elements (trees, cars, people, etc.);

·LED lighting strips used to indicate energy flows between the participants;

·LED lights on the buildings representing various types of activities (energy generation, in-house activity, and energy storage);

·Model control unit;

·Necessary cables and connectors;

·Model base.

Actors’ interactions with the model would, as mentioned above, take place through table computers, which would be fitted out with appropriate software solutions, such as the oTree experiment software[24].There would be one additional control/server unit utilised for initiating experimental and other interactions as well as for managing and monitoring the model, likely using a browser-based interface.Model functions would be controlled by the researchers through the control unit, which should be based on the Raspberry Pi system.The control software is planned to be developed using the Python coding language, which will improve the cross-compatibility with the experiment software, which is also based on Python[24].

It is intended for the model to be able to operate in several different modes, starting with a demonstration mode similar to the operation of the “old” model, and extended to visualisations of actions undertaken by people interacting with the model.These visualisations could include, for example, visualisations of actors’investment decisions and visualisations of energy trading processes.Table 2 is an overview of the model’s visual outputs and their roles in the demonstration mode as well as in two potential interactive scenarios—one scenario based on players making real-time investment decisions for a hypothetical decentralised energy system and the second scenario simulating energy trading.

Fig.4.Sketch of the planned model (own presentation).

Table 2: List of LED indicators on the new crowd energy model concept (own presentation)

The two potential interactive scenarios would allow the actors to engage in the behaviour related to their real-life decision, making in the context of a cooperative, decentralised energy system.The first scenario would be related to their initial decisions to invest in such a system, while the second scenario would represent the daily trading of(excess) energy within the system, with actors taking the roles of energy sellers and buyers and being able to agree trades for a set price.Both of these decisions could be shown in real time on the model, adding a degree of interactivity compared to standard social sciences games and experiments.

In Figs.5 (a) and (b), the authors illustrate how the model could visualize actors’ decisions for the investment and energy trade scenarios, showing a random situation.In both cases, there are eight actors taking part in the scenarios, for the energy trade scenario they are separated into two groups of four sellers and buyers.

Fig.5.Model visualization for (a) investment decision scenario and (b) energy trade scenario (own presentation).

In this situation, all actors were asked to make an investment decision with an initial endowment of 100 units,being able to choose any amount between 0 and 100.After all decisions were made, the model would indicate the invested amount by lighting a number of LEDs corresponding to the ratio between the invested amount and the full endowment, rounded up, with each LED representing 20 percent (for example, player E in Fig.5 (a) invested 50 units, which is 50 percent of the full endowment, resulting in 3 LEDs being lit).

In Fig.5 (b), actors were separated in energy sellers (A, B, G, and H) and buyers (C, D, E, and F); actors A and E, B and C, and D and H were able to agree an energy trade, while actors F and G did not engage in trading.As soon as any of these trades were agreed, the model would indicate the trade by lighting the LED strip connecting the houses representing the trading players in a sequence starting from the seller and ending with the buyer.Compared to that in Fig.5 (a), this would happen in real time—as soon as a trade would be agreed, the model would indicate it.

4. Outlook

The use of a physical interactive crowd energy model combined with information gathering and education through experiments was considered by the authors as a potentially highly fruitful approach for the further development of the crowd energy concept, particularly towards practical, real-life applications.Following the conceptual and planning work presented in this paper, the next step will be the physical construction of an interactive model and the programming of the software necessary for the implementation of the interactive scenarios and/or experiments.

In the course of this development, it would be useful to retain an element of modularity to the model—both in terms of its physical shape and its virtual capabilities.By utilizing a modular concept, the researchers could adapt the model in order to represent larger or smaller communities, while the software could be extended towards additional functionality—for example, the modelling of additional environmental and contextual factors, such as weather or transmission losses in energy grids, or the ability to implement and run completely new scenarios, such as multi-action experiments or the simulation of the full development process of a crowdtype energy system.

Further, the software could be developed with an eye towards “conventional” economics experiments performed with a larger number of actors in a classroom or laboratory setting[22].By including this particular functionality, researchers could perform larger-scale experiments yielding data that is more robust; furthermore, by separating the software from the model, these experiments could be performed either with or without framing, with the physical model serving as an (optional) framing device.

Acknowledgements

The authors would like to thank the Canton of Fribourg, Switzerland, for the support through the smart living lab project at the University of Fribourg.