Available online at www.sciencedirect.

com

ScienceDirect
Available online at www.sciencedirect.com

ScienceDirect
Procedia Manufacturing 00 (2019) 000–000
www.elsevier.com/locate/procedia

Procedia Manufacturing 51 (2020) 1117–1124

30th International Conference on Flexible Automation and Intelligent Manufacturing (FAIM2021)

15-18 June 2021, Athens, Greece.

Architecture Model for a Holistic and Interoperable Digital Energy

Management Platform
Pedro P. Sennaa,b,*, António H. Almeidaa, Ana C. Barrosa, Ricardo J. Bessaa, Américo L.
Azevedoa,b
a
INESC
INESC TEC - Instituto de Engenharia de Sistemas e Computadores, Tecnologia e Ciência, Rua Dr. Roberto Frias, Porto 4200-465, Portugal
b
FEUP - Faculdade de Engenharia da Universidade do Porto, Rua Dr. Roberto Frias, Porto 4200-465, Portugal

* Corresponding author. Tel.: +351-22-209-4398. E-mail address: [email protected]

Abstract

The modern digital era is characterized by a plethora of emerging technologies, methodologies and techniques that are employed in the
manufacturing industries with intent to improve productivity, to optimize processes and to reduce operational costs. Yet, algorithms and
methodological approaches for improvement of energy consumption and environmental impact are not integrated with the current operational
and planning tools used by manufacturing companies. One possible reason for this is the difficulty in bridging the gap between the most advanced
energy related ICT tools, developed within the scope of the industry 4.0 era, and the legacy systems that support most manufacturing operational
and planning processes. Consequently, this paper proposes a conceptual architecture model for a digital energy management platform, which is
comprised of an IIoT-based platform, strongly supported by energy digital twin for interoperability and integrated with AI-based energy data-
driven services. This conceptual architecture model enables companies to analyse their energy consumption behaviour, which allows for the
understanding of the synergies among the variables that affect the energy demand, and to integrate this energy intelligence with their legacy
systems in order to achieve a more sustainable energy demand.

© 2020The
© 2020 TheAuthors.
Authors. Published
Published by Elsevier
by Elsevier Ltd. Ltd.
This
This isisan
anopen
openaccess
access article
article under
under the BY-NC-ND
the CC CC BY-NC-ND licenselicense (https://creativecommons.org/licenses/by-nc-nd/4.0/)
https://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review underresponsibility
Peer-review under responsibility ofof
thethe scientific
scientific committee
committee of FAIM
of the the FAIM 2021.
2020.

Keywords: Energy efficiency; IIoT Platform; Architecture Model; Energy data driven services; Energy Digital Twin

1. Introduction reduction of EC by 2020 and 32.5% reduction by 2030

compared to 2007 baseline projections [3]. To achieve such
Climate change urges nations worldwide to take strong and scenario, the European Commission has launched the “Energy
assertive actions towards the mitigation of global warming and Efficiency Plan”, seeking to promote industrial energy
its impact in citizens. The current trends in energy consumption efficiency by establishing strict energy requirements for
(EC) and related emissions will lead us to a bleak future, industrial machinery and equipment [3].
especially when considering the recent data regarding historic The manufacturing industry is amongst the biggest
levels of emissions, i.e. the rise of energy-related CO2 consumers of primary energy, representing 31% of the primary
emissions to reach the all-time high levels of 33.1Gt in 2018 EC, as well as amongst the largest CO2 emitters [4]. This
[1]. On this note, the European Union (EU) is undertaking an industrial sector is capable of increasing the energy efficiency
ambitious roadmap, called “European Green Deal”, to become from 18% to 26%, and of reducing CO2 emissions from 19%
the first climate-neutral continent by 2050, as well as to slow to 32% [4], with unrealized long-term improvement
down global warming and mitigate its effects [2]. Within this capabilities of up to 60% energy-efficiency ratings in the long-
initiative, the EU has made the commitment of achieving a 20%
2351-9789 © 2020 The Authors. Published by Elsevier Ltd.
2351-9789
This is an©open
2020 access
The Authors. Published
article under bytheElsevier Ltd.
CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
This is an openunder
Peer-review access article under the CC
responsibility BY-NC-ND
of the license
scientific https://creativecommons.org/licenses/by-nc-nd/4.0/)
committee of the FAIM 2021.
Peer-review under responsibility of the scientific committee of the FAIM 2020.
10.1016/j.promfg.2020.10.157
1118 Pedro P. Senna et al. / Procedia Manufacturing 51 (2020) 1117–1124
Senna et al./ Procedia Manufacturing 00 (2021) 000–000

term due to its extensive size and meaningful impact in the approach of Industrial Internet-of-Things (IIoT), Digital Twin
global energy efficiency [5]. and i4.0 enabling technologies, with energy data driven services
The use of digital technologies and operations research and data ubiquity, on an interoperable digital platform targeting
methodologies for improving industrial efficiency has had energy management and efficiency. Each of these aspects is
increasing efforts since the advent of the industry 4.0 paradigm, further described on the following subsections.
born from the German Plattform Industrie 4.0 set out in 2013
[6]. The industry 4.0 (i4.0) paradigm presents a set of 2.1. Interoperable Digital Platforms for Energy Efficiency
contemporary technological advances that integrate physical
objects, virtual models and services, as well as coordination In technical terms, digital platforms are presented in a multi-
efforts [7], that present three major benefits to the layer setup that ranges from web portal and business
manufacturing industry: reduction of operational costs; interactions to the software tools and data management layers,
increase in production efficiency; and additional revenues from as well as the communication layer that intertwines the entire
the servitisation of products [8]. In what concerns ICTs used ecosystem [18]. The foundation is usually constituted of a
for energy-efficiency purposes, the EU Strategic Energy semantic infrastructure, with ontologies and inference rules,
Technology Plan (SET) aims both to propose an interoperable which is combined with a database to enable ecosystem data
reference architecture, significantly improving the electricity management capabilities and interoperability among different
consumption management and reduction of costs while assets. The services are applied in a plug-and-play fashion into
increasing the penetration of user-friendly tools [9], as well as the ecosystem data manager, and further integrated into the web
to promote changes in the energy mix towards increasing use portal through a gateway orchestrator [19].
of renewable and sustainable alternatives following the efforts Digital platforms have notorious implications for the
put forward since the Paris Agreement [10]. economy given their inherent separation between physical &
Clearly, there is a need for stronger efforts to integrate all virtual assets and the value created by them [20]. In order to
the developed digital solutions to achieve a smart, holistic, enhance the collaborative potential and the widespread
scalable and green manufacturing industry environment based dissemination of information & data, a holistic value network
on energy efficiency and cleaner environmental footprint. perspective must be adopted, which expands from the simple
Therefore, this paper proposes a conceptual architecture model matchmaking of manufacturing resources to a manufacturing
for an interoperable and holistic energy management digital ecosystem network. With such approach, the digital platform
platform for manufacturing plants. This conceptual architecture can combine tools and ecosystems, through which sustainable
model will leverage energy data interoperability, capable of use of resources may be achieved [20]. The drivers to the
enriching existing legacy systems with intelligent energy development and implementation of digital platforms for the
perspective, thus, covering the gaps previously identified. manufacturing industries, also known as digital platform
Moreover, it will promote flexibility, interoperability and affordances, usually pertain the bridge between technological
secure communication, in order to be the supporting ground for availability & maturity, and the human-machine interaction
virtual and real-time simulation & optimization of industrial capabilities. They have been identified as follows [20]:
processes towards leapfrog improvements in energy efficiency.
Its modularity perspective allows for numerous combinations • The need for enabling flexibility (from modular-driven
of services regarding the manufacturing sector context and the services and sharing economy to user-IO and role changing
individual characteristics of the industrial units. capabilities) [21];
• Matchmaking between different attributes [22];
2. Literature Review • Scalability capabilities and cross-sectoral reach extensions
[23];
The conceptual architecture model proposed in this research • Management of all transactions in the ecosystem [24];
is an abstract framework for multi-case instances, based on • Building trust through high-security measurements and full
multiple combinations of implementation configurations, and information transparency among players [18];
designed for empirical applications and solutions of • Support the engagement of the network with the aim of
manufacturing companies [11]. In the reference models’ establishing a strong community [25].
literature there are multiple instances targeting energy-related
aspects, such as those pertaining electrical energy systems in Despite conceptual developments and the availability of
Smart Grid environments [12], green buildings’ construction solutions, a flexible, secure and interoperable digital platform
[13], and cross-flow turbines [14]. Most of these reference has not yet been developed with the clear purpose to assess the
models present sufficiently complete frameworks to address energy-efficiency of industrial complexes through smart
specific problematics in energy manufacturing domains. sensoring and simulation & optimization capabilities.
Additionally, there are multiple reference models for the digital
technologies, such as the digital twin architecture reference 2.2. Energy Related Data Ubiquity
model for cloud-based cyber-physical systems [15], Internet-
of-Things (IoT) middleware [16], and hybrid reference model In order to achieve decreased EC, energy efficiency and
matching with virtual reference feedback among smart carbon footprint reduction, companies can implement changes
controllers [17]. Yet, it is clear that there is a lack of a on manufacturing processes, which can range from renewable
conceptual reference architecture model for a combined energy generation, energy transformation techniques and
 Pedro P. Senna et al. / Procedia Manufacturing 51 (2020) 1117–1124 1119
Author name / Procedia Manufacturing 00 (2019) 000–000

ubiquitous sensing, to smart machinery control and in several manufacturing processes mainly due to difficulties in
development of energy/comfort systems (light, air, heating and obtaining shop floor calibrated measurements [33]. The state-
cooling, windows and doors) [26]. One way of reducing the EC of-the-art data-driven services are mainly limited to:
is by increasing the efficiency of existing machinery, i.e. belt product/key parts EC information monitoring and forecasting
conveyors energy efficiency increased through variable speed to estimate indicators (e.g., EC proportion of each key part) and
control [27]. Another way is reusing the wasted energy inside integrate them in green material selection [34]; evaluate daily,
the plant. Heat waste energy can be transformed into another week and yearly trends of EC [35]; and, cloud-based services
form of energy, generally electricity. These techniques deal for EC estimation in machining operations, including
with the reduction of EC in individual plants separately, but a information from human operators and process plan [36]. In
collective approach targeting multiple industrial units can be terms of process control and energy optimization, different
achieved through the deployment of industrial symbiotic studies explored model-driven energy optimization techniques,
techniques, which involve the exchange of materials, energy, such as integer programming for peak load reduction in steel-
water and by-products [28]. plants [37] and robust optimization for real-time control of a
Due to the powerful embedded devices available today, the steel powder manufacturing process under varying prices [38].
computing power is being brought closer to sensors and many The advent of IoT technology offers technical conditions for
tasks previously executed on desktop computers, SCADA data-driven approaches for energy optimization, where, for
(Supervisory Control and Data Acquisition) systems or cloud instance, each process can be represented by a process
are being moved to smarter Internet-of-Things (IoT) devices. input/output matrix [39] or a virtual battery [40]. This approach
This strategy allows data pre-processing and filtering, reducing does not require a full modelling of the process equations since
the amount of storage needed for applications, which is very its understanding is made in real-time through data and can be
important due to the huge amounts of data generated by IoT combined with a digital twin manufacturing system that
devices and the data transmission through the network [29]. In performs simulations and decisions for different energy-saving
addition, this strategy reduces the response time in control purposes in a virtual space, e.g. event-driven energy-saving
applications, enabling real-time control, converting the IoT decision method that switch machines to sleep mode with
sensing devices into smarter components usually referred as minimum effect on the system throughputs [41].
edge computing devices [30]. These devices are commonly Recent advances in machine and deep learning techniques,
connected to the cloud for data storage and updating of the combined with the increasing monitoring and simulation
machine learning algorithms used to forecast and control. Most capability offers by the i4.0 paradigm, enables different use
modern architectures allow for online update of the models cases for data-driven services, namely:
while the IoT device is in operation.
Energy measurement and forecasting opens up the • Detection of abnormal consumption patterns and root-cause
possibility to apply smart control to energy and production analysis to find energy efficiency actions.
systems. Production line measurements and forecasts are • Benchmarking either between similar manufacturing
combined in the literature with other data to reduce the EC and industries or between “perfect” (in terms of energy
CO2 footprint of the manufacturing process. Nevertheless, consumption) digital twin and real factory.
seldom research has attempted to combine all the different • Data-driven modelling of industrial processes (including
approaches. Energy forecasting of complete manufacturing energy consumption) and inclusion in digital twin
lines or buildings have much greater potential, as it allows for simulation as proxies to produce recommendations for
a global optimization that includes not only the EC or the cost, optimal tuning scenarios.
but also other variables. Although several works in commercial • Estimate process flexibility for power system ancillary
areas and energy systems in buildings in general are available, services provision (see [42] for taxonomy and more details
there are few works that tackle energy reduction at an industrial about flexibility and ancillary services products), e.g.
level. predictive consumption curtailment to offer peak load
reduction to transmission system operators.
2.3. Energy Data Driven Services • Predictive control for energy optimization of continuous and
energy-intensive processes.
Most industrial facilities are increasing their energy
efficiency by implementing ISO 50001 approved methods [31], For the last two services, some promising techniques from
aiming to decrease energy costs and greenhouse gas emissions. the state-of-the-art can be exploited. For flexibility modelling,
This standardized practice provides monitoring and awareness the representation proposed in O’Connell et al. [43] for
of EC to human decision-makers, but does not provide flexibility (i.e., power versus duration curves), and modelled
prescriptive analysis and/or autonomous process control. with ARMAX (Auto-Regressive Moving Average with
Moreover, basic descriptive analytics (e.g., linear regression) eXogeneous Input) models, can be extended to heat or coal
might lead to inadequate and simplified models of energy based industrial processes. With this modelling approach, the
efficiency monitoring [32], since they do not consider different industrial consumer can estimate the power flexibility and
factors, such as financial costs, size of the installation, present offers in the electricity market (direct participation or
consumption of different raw materials, implementation of represented by a market aggregator). For predictive energy
other energy efficiency actions, etc. It is also important to optimization of continuous processes, a promising approach is
underline that the level of awareness regarding EC remains low to combine supervised learning (development of functional
1120 Pedro P. Senna et al. / Procedia Manufacturing 51 (2020) 1117–1124
Senna et al./ Procedia Manufacturing 00 (2021) 000–000

relations between state and control variables for different types comprehend the requirements of manufacturing companies
of processes and forecast demand in energy intensive flows) across industrial sectors regarding energetic aspects of
and reinforcement learning (self-learning – e.g. through production lines. Based on this research methodology, a
Proximal Policy Optimization methodology) for emulating the conceptual architecture model for Digital Energy Management
physical environment from data. This approach was applied to Platform was developed, grounded on six main perspectives,
control the variable-speed pump operation in wastewater tanks towards bringing energy efficiency into the manufacturing
(see Filipe et al. [44,45] for more details), and can be extended industry. These perspectives are further defined:
to other industrial processes if data from sensors that
characterize the system state are available. • Support a digital continuity paradigm towards an optimal
energy management in a holistic way along the entire value
2.4. Digital Twin for Predictive Energy Management chain.
• Provide a holistic and ubiquitous energetic information for
Usually, IoT solutions applied to the energy domain can be full energy demand and consumption characterization
structured in three logical layers [46]: amidst the different levels of granularity of complex
manufacturing systems. This is achieved through an open,
• Wireless sensor networks, which detect the signal coming standardized and interoperable energy management
from the production line; framework, which is fully synchronized with real industrial
• Registers (accumulators), where the energy consumption is environments.
measured and stored separately for process; • Enable energy data interoperability with manufacturing
• External servers and software to monitor the energy legacy system to include the energy and sustainable
consumption in the factory. perspectives within the regular planning and optimization
tools. Enriching manufacturing and enterprise legacy
An example of such solutions are the distributed real-time systems with energy intelligence will enable manufacturing
energy monitoring system, build upon the wireless sensor companies to take decision also considering energy
network, aimed at supporting machine scheme selection and efficiency, maximization of energy use from RES and
energy quota allocation in the shop floor [47]. On the top of the reduce costs with energy.
data acquisition layer, a digital twin (DT) can be applied to • Understand synergies between endogenous and exogenous
model the characteristics of the real networked system. This variables and its impact on energy consumption. Using
concept is further extended to the shop-floor, thus achieving the Artificial Intelligence as a tool to identify patterns on energy
digital twin shop-floor (DTS) [48], which aims at merging the demand and generate intelligence on energy consumption
physical and virtual spaces to optimize the current production behavior, it is expected to support manufacturing companies
activities. In DTS, every element in Physical Shop-floor in understanding multi-variable cause-effect relations with
(human, the equipment, the material and the environment) has impact on energy demand and consumption
its digital representation [49]. A DTS allows the integration of • Expand Human worker cognitive capabilities to take the
data provided by both the physical and the virtual space to most sustainable option in their decision-making process.
optimise performances of the system and manage energy Using proper human centric technologies, it is possible to
consumption. These technologies allow the management of provide humans with energy intelligence that can rule the
various system parameters concurrently, making them suitable decision-making along the entire value chain.
for managing complex systems. Applications in this field • Optimize energy demand and promote reduction of costs
allowed to address energy consumption monitoring, data and of environmental impact through smart industrial
analysis, energy efficiency improvement, environmental processes’ configuration and parametrization. The key to
impact reduction, productivity increase and cost reduction [50]. fulfill this purpose is a set of innovative data-driven energy
services, capable to prescribe industrial processes
3. Architecture Model for a Digital Energy Management configuration and assets parameterization towards energy
Platform efficiency, environmental impact mitigation and energy cost
reduction.
The conceptual architecture model for Digital Energy
Management Platform was designed through a combination of
literature review and non-structured interviews. The literature
review considered both scientific and grey literature, and aimed
at understanding the current state-of-the-art of architectures for
digital platforms targeting energy outcomes by combining ICT
techniques and tools into legacy systems and smart machinery
within manufacturing companies. The non-structured
interviews were conducted with manufacturing companies
from five European countries (Germany, Spain, Portugal, Italy
and Sweden), and among different manufacturing sectors:
testing and experimentation facilities, automotive parts, iron
Fig. 1 Architecture Model
foundry, textile and chemical. These interviews sought to
 Pedro P. Senna et al. / Procedia Manufacturing 51 (2020) 1117–1124 1121
Author name / Procedia Manufacturing 00 (2019) 000–000

In order to improve their energy efficiency, companies may strict guidelines, mitigate the data flow with the platform,
apply managing and monitoring activities on their factory, reduce data latency and guarantee stronger capacity for agile
product and process designs in an integrated way. Typically, detection of bad behaviours with quick reaction. This approach
this incurs in optimizing the parameterization and configuration will tremendously ease the development of high-level
according to the specificities of the task in production, such as algorithms, like energy forecasting or digital twins, which do
the product composition and the environmental characteristics not need to deal with low-level details.
(e.g. weather conditions). Furthermore, it looks to put forward
the simulation tools to explore and validate potential industrial 3.2. Pillar 2 – Human-Machine Interaction
and internal processes’ symbiosis, while aiming to prioritize
retrofitting options, which focus on decreasing waste, An important perspective of i4.0 is the empowerment of
leveraging the energy circular economy and maximizing the employee’s cognitive capabilities towards a more intelligent
return on investment (ROI) when updating or replacing the and complex decision-making process, taking advantage of
assets and machines. enabling technologies explored and developed within the scope
On this note, AI based data-driven services, which are fully of the i4.0. Thus, the main objective of this pillar is to study and
available and accessible to employees, provide richer assess enabling technologies, such as Augmented and Virtual
information on energy demand patterns and potential root reality, 3D simulation environments and other technologies
causes, as well as prescribe improvement actions. In this way, capable of: (i) presenting relevant energy-related information at
it will be possible to make effective and sustainable data-driven the right moment to decision makers in a more human-centric
decisions, as well as to anticipate the future impact of these way and (ii) enabling the power of data driven services (Pillar
decisions, especially in foreseeing potential energy savings and 4) for humans and guaranteeing that these services can be used
energy costs. The conceptual reference architecture model for a interactively by decision makers according to their needs.
digital energy management platform will bridge two main These will take advantage of the green and energy efficient
concepts: (i) the supporting information & communication framework, which will be explored in Pillar 3, in order to
technologies (ICTs) emerging within the i4.0 context; and, (ii) exploit a complete and innovative way of presenting
the enterprise/manufacturing systems and external systems information for different decision makers along the complex
specific from the energy industry (Fig. 1). On a broad manufacturing system levels (from the shop floor to the factory
perspective, the conceptual reference architecture model aims and supply chain perspectives). It seeks to build rich, modular
to integrate these perspectives to propose a holistic and and interoperable energy information models, capable of
distributed Energy Management Expert suitable to be integrating and empowering existing digital twin models (e.g.
implemented in SMEs & mid-caps, in big companies and in production, product and building models) to support the
multinationals. creation of immersive environments. In these immersive
environments, decision makers can enhance their awareness
3.1. Pillar 1 – Factory Driver IO about real-time EC, efficiency and quality, comparisons with
their digital shadow behaviour, receive alerts of potential
The main idea of the Factory Driver IO is to create an problems and a prognosis about the root causes and points of
abstraction layer that makes interaction with the production attention. Workers can also benefit from real-time data and
system easier from high-level applications. Considering the information exchange, which will enhance productivity, safety
huge quantity of energy-related data that is expected to be and complexity, while providing security and reliable decision-
generated, it becomes critical to use emergent data management making strategies.
and data fusion techniques capable of pre-processing them and
guaranteeing its accuracy and quality, which is achieved 3.3. Pillar 3 – Energy Data Modelling & Standardization
through edge computing techniques. A set of edge devices,
based on Graphic Processing Unit (GPU) System-on-Chip Given its role in increasing energy-efficiency, the semantic
(SoC) will be used given their seamless development, increased interoperability based on standardized ontologies brings a
reliability and reduced costs & power consumption. These use significant value proposition for the complete integration of IoT
industrial protocols to connect to sensors, actuators, robots and energy-related information in the manufacturing domain. This
other machines, and the rest of applications, generating a evolution overcomes the limitations imposed by complex
unified interface. The edge devices will be installed in the solutions of existing systems architecture, where different
production line, being capable of supporting most common manufacturing systems already exist, as well as optimization
industrial protocols such as Profibus, Profinet and Modbus. The and planning tools, sometimes from different providers or
architecture supports different field bus protocol in each edge belonging to monolithic solutions. Yet, it requires more, and
device and even more than one protocol per edge device, better, energy information to achieve greater green- and energy-
providing flexible and hybrid communication networks. All efficient performances. The standardization of solutions and
low-level configurations are encapsulated in the edge device, data models, and the interoperability across IoT services and
which provides access to the production system always with the datasets through reference ontologies (developed in the scope
same data type and units, regardless of the type of sensor or of this pillar), are key concepts towards the effective adoption
communication protocol used at shop floor level. Thus, it of new ICT technologies and empowerment of existing ones for
becomes possible to distribute the data processing while energy management. This pillar is executed with the main
keeping local control of processes and machines according to objective of enabling standardization and acceptance of new
1122 Pedro P. Senna et al. / Procedia Manufacturing 51 (2020) 1117–1124
Senna et al./ Procedia Manufacturing 00 (2021) 000–000

data models, both in the research and industrial domains, model explores the development of intelligence (pillar 4)
capable of leveraging new green- and energy-efficient capable of enriching existing planning and management
certifications, labelling and regulation procedures. enterprise tools, through a seamless interoperable mechanism
based on energy digital twin models (pillar 3), with a holistic
3.4. Pillar 4 – Data-driven Services view on energy quality and demand (pillar 1), while exploring
human-centric technology to enhance human decision makers’
The main motivation behind data-driven services is that cognitive capabilities (pillar 2).
physical modelling of different industrial processes might be The main challenge of the Energy Management Expert
complex and/or expensive to obtain and it does not fully boost System is related to the need to take advantage of intelligence
the replication potential of the Energy Management Expert related to energy demand and quality, and integrate it with the
System. Thus, the data collected from energy- and process- enterprise and manufacturing systems, towards a more efficient
related smart sensors, which is aggregated and pre-processed in EC profile. This will be achieved through four different
the IIoT Platform, and potentially combined with data collected approaches:
from digital twin simulation models, is exploited to offer data-
driven energy services supported by: (i) big data solutions and • Energy Intelligence for Legacy Systems, which focuses on
(ii) state-of-the-art AI modelling framework from the AI4EU energy-related information interoperability through
platform. The following groups of services can be developed different perspectives of the production system
and integrated with the Energy Management Expert System: management, aimed at achieving optimal energy saving
potentials, energy- and eco-efficiency (by using energy
• Predictive analytics (or prognosis): Forecast optimized intelligence integrated into scheduling tools – Green
process operations, such as finding functional relations Scheduling), while maintaining business targets (use of
between state and control variables of industrial processes ERP/MES, production planning tools, enterprise
with influence in energy consumption. Modelling management systems, product design & management tools).
uncertainty (e.g., unplanned events and other perturbations) • Life Cycle Assessment and Cost Analysis (LCA/LCC)
will be a key requirement to assess feasibility of process tools, which are integrated in the digital energy platform, to
scheduling. These business-driven analytics can also support the assessment of the best energy management
provide real-time analytics and simulation (e.g., embedded decision considering past performances and forecasted
in the digital twin) of KPIs related to energy consumption. energy behaviour, all of which are obtained through the
• Prescriptive analytics: Self-control of energy consumption industrial pilot cases. Having both a company-specific and
in continuous industrial processes that can operate with a value chain-oriented approach, this integration will
minimum control and monitoring requirements and are provide sustainability values enabling to compute the
easily traceable and adaptive to process changes. Root cause different scenarios related to energy usage and provide
analysis of abnormal consumption partners and cause-effect optimization capabilities.
analysis for identification of energy-efficiency actions are • Simulation and Optimization mechanisms supported on the
also envisioned services. These services can work both in energy digital twin models developed. Within 3D and
centralized and decentralized decision-making frameworks. immersive environments, it will be possible to create and
validate hypothesis not only related to the company’s
The main challenge is to produce a modular framework for layout, product characteristics and industrial processes
energy services, which can be adjusted and scaled to model configuration, but also to their synergies and impact on
systems that are continuously changing while covering different energy perspectives.
different end-user requirements. Moreover, a human-centred • Self-Automation and Control capabilities, achieved through
perspective on the interface design must be considered in order an energy control station, where managers from the
to enable humans to best exploit the available data, make better industrial pilot cases can define very simple business
decisions, focus on core tasks and be able to perform better in guidelines, based on IF/THEN/ELSE logic capable to be
an increasingly networked work environment, therefore being deployed in the edge computing infrastructure for cyber-
integrated with the Factory Driver I/O of Pillar 1 and with the physical systems (CPS) control towards a more energy-
Human-Machine Interaction of Pillar 2. efficient performance.

3.5. Energy Management Expert System 4. Illustrative Use Case

Considering the complexity of today’s manufacturing To validate the proposed conceptual architecture model, a
systems, it is clear that decision makers at all levels of the theoretical case study was designed based on the case research
organization and value chain need expert systems for real-time protocol from Yin [51]. For this purpose, a Portuguese iron
decision-making on a broader, holistic overview. Therefore, for foundry manufacturing company was selected, which produces
emergent topics with complex settings (i.e. high number of rough and pre-machined parts in nodular and grey iron for the
variables, multiple trade-off scenarios, synergetic nature and automotive industry. This theoretical case study focuses in the
requirements, and impact on business drivers), such as the operational optimization of the brazing furnaces used for
energy efficiency in manufacturing, this represent a huge welding, through real-time energy consumption prediction and
advantage. Therefore, the conceptual reference architecture digital twin. The furnaces’ overheating processes demand a
 Pedro P. Senna et al. / Procedia Manufacturing 51 (2020) 1117–1124 1123
Author name / Procedia Manufacturing 00 (2019) 000–000

massive EC to achieve a temperature of 1500ºC, on average, to applications and solutions for combined ICT and enterprise
conduct their melting tasks. This temperature set point depends systems’ approach towards enhancement of energy efficiency
on a set of variables, such as the final product characteristics in manufacturing industries. Thus, the proposed conceptual
and the composition of the raw material used, and their correct architecture model is a scalable, interoperable and holistic
calculations are considered critical to achieve balance between architecture model capable of collecting and pre-processing all
quality and EC. It is, therefore, a complex set of processes, energy-related data, taking advantage of innovative CPS
which is currently conducted by well-trained and experienced connected along the shop floor, enabling access to advanced
workforce. Nevertheless, to achieve optimized energy AI-based data driven services for generation of energy
efficiency, it is necessary to implement intelligent systems to intelligence, which can be integrated in manufacturing
monitor and control the furnaces by considering the companies’ legacy systems.
endogenous and exogenous variables of this complex system. Future works to build on this research should focus on
To overcome this limitation, a digital platform is to be validating the proposed conceptual architecture model through
deployed, where endogenous and exogenous variables can be an established multi-case research design, as well as on
collected and pre-processed in real-time to feed AI based data- developing solutions based on such architecture to be
driven services. The result is grounded knowledge generation implemented in manufacturing companies. The multi-case
to allow for controlled optimized decision-making, thus studies are to be conducted within the manufacturing
enabling energy efficiency. Digital twins, each exclusive to a companies already interviewed for this research, as depicted in
separate furnace, are to be defined according to the machine’s section 3, with a cross-sectorial approach to ensure full
specifications, which will consider control mechanisms, validation of the proposed conceptual architecture model. This
physical-, electro-magnetic- and thermal-curve/behavior, as instantiation within industrial companies is expected to provide
well as EC patterns. This digital model will be continuously energy efficiency gains, which lead to life cycle cost reductions
updated with rich information from the algorithms, which will of production lines and lower environmental impact.
be pre-processing data from the furnaces, at the edge level, and From an academic perspective, the focus should be on
publishing information to the cloud in real-time. In parallel, AI providing interoperability capabilities among enabling
based data-driven services will be reading and correlating this technologies and legacy systems that easily integrate with this
information to generate knowledge that can enrich even more reference architecture model.
these digital models so that multivariable nonlinear control of
the overheating process can be performed. Acknowledgements
By combining all this information within proper simulation
technologies, it is possible to provide decision-makers with the This work is financed by National Funds through the
capability to optimize production planning and configuration, Portuguese funding agency, FCT - Fundação para a Ciência e
and the furnace parameterization, towards energy efficiency a Tecnologia, within project UIDB/50014/2020.
while keeping production rate within quality parameters and
reducing environmental footprint. This considers all the References
production perspectives (e.g. final product characteristics, raw
materials, furnace’s status and behavior, weather/temperature [1] International Energy Agency (IEA, 2019). IEA @ COP25. Available at:
https://www.iea.org/reports/iea-cop25 Accessed on: 10.01.2020
conditions). A simulation engine will use this results to enable [2] Von der Leyen (2019). The European Green Deal: EU’s Response to
operators to monitor the current status of the furnaces and plan Climate Change. European External Action Service. Available at:
the overheating process considering the production scheduling. https://eeas.europa.eu/headquarters/headquarters-
homepage/72184/european-green-deal-eus-response-climate-change_en
5. Conclusions Accessed on: 10.01.2020
[3] European Commission (2018). The Energy Efficiency Directive
(2012/27/EU) and The Amending Directive (2018/2002). Available at:
In light of the challenges set forth by the European https://ec.europa.eu/energy/en/topics/energy-efficiency/targets-directive-
Commission and the current global landscape regarding EC, and-rules/energy-efficiency-directive#content-heading-0 Accessed on:
environmental impact reduction and the role of the 05.01.2020
manufacturing industry, this research has presented a [4] International Energy Agency (IEA). (2007). Tracking Industrial, Energy
Efficiency and CO2 Emissions. Available at:
conceptual architecture model for a digital energy management http://www.iea.org/textbase/nppdf/free/2007/tracking_emissions.pdf.
platform. Such conceptual architecture model is supported by Accessed on: 05.01.2020
four well-designed concepts surrounding data ubiquity, data [5] International Energy Agency (IEA). (2012). World Energy Outlook 2012,
driven services, digital interoperable platforms and digital twin World energy outlook special report. International Energy Agency, Paris,
modelling for energy efficiency and environmental impact France.
[6] European Commission (2017). Implementation of an Industry 4.0 Strategy
reduction. – The German Plattform Industrie 4.0. Digital Single Market. Available at:
Moreover, the conceptual architecture model is comprised https://ec.europa.eu/digital-single-market/en/blog/implementation-
of four pillars that promote an energy management expert industry-40-strategy-german-plattform-industrie-40 Accessed on:
system that allows for modular applications of energy data 10.05.2019.
driven services considering human-machine interactions [7] Drath, R., & Horch, A. (2014). Industrie 4.0: Hit or hype?[industry forum].
IEEE industrial electronics magazine, 8(2), 56-58.
through a factory driver i/o, and with full documentation in [8] Schumacher, A., Erol, S., & Sihn, W. (2016). A maturity model for
order to establish common framework. The proposed assessing Industry 4.0 readiness and maturity of manufacturing enterprises.
conceptual framework can be further extended towards full-set Procedia Cirp, 52, 161-166.
1124 Pedro P. Senna et al. / Procedia Manufacturing 51 (2020) 1117–1124
Senna et al./ Procedia Manufacturing 00 (2021) 000–000

[9] European Commission (2018). SET Plan delivering results: The [29] Molanes, R. F., Amarasinghe, K., Rodriguez-Andina, J., & Manic, M.
implementaiton Plans - Research & Innovation enabling the EU's energy (2018). Deep learning and reconfigurable platforms in the Internet of
transition. Publication prepared jointly by European Commission's Things: Challenges and opportunities in algorithms and hardware. IEEE
Directorates-General for Energy, Research & Innovation and Joint industrial electronics magazine, 12(2), 36-49.
Research Centre. Publications of the European Union - Project number [30] Yu, W., Liang, F., He, X., Hatcher, W. G., Lu, C., Lin, J., & Yang, X.
2018.5779. Available at: (2017). A survey on the edge computing for the Internet of Things. IEEE
https://setis.ec.europa.eu/sites/default/files/setis%20reports/setplan_delive access, 6, 6900-6919.
ring_results_2018.pdf Accessed on: 05.01.2020. [31] Kluczek, A., Olszewski, P. (2017). Energy audits in industrial processes.
[10] International Energy Agency (2020). Data and Statistics - Total Primary Journal of Cleaner Production, 142, 3437-3453.
Energy Supply by source, European Union - 28 1990-2017. Available at: [32] Nakhodov, V., Baskys, A., Skeie, N. O., Pfeiffer, C. F., Dmytro, I. (2016).
https://www.iea.org/data-and- Selection methodology of energy consumption model based on data
statistics?country=EU28&fuel=Energy%20supply&indicator=Total%20pr envelopment analysis. Electrical, Control and Communication
imary%20energy%20supply%20(TPES)%20by%20source Engineering, 11(1), 5-12.
[11] MacKenzie, C. M., Laskey, K., McCabe, F., Brown, P. F., Metz, R., & [33] Owodunni, O. (2017). Awareness of energy consumption in
Hamilton, B. A. (2006). Reference model for service oriented architecture manufacturing processes. Procedia Manufacturing, 8, 152-159.
1.0. OASIS standard, 12(S 18). [34] Tao, F., Bi, L. N., Zuo, Y., & Nee, A. Y. C. (2016). (Tao 2016). CIRP
[12] Cespedes, R. (2012, September). A reference model for the electrical Annals, 65(1), 9-12.
energy system based on Smart Grids. In 2012 Sixth IEEE/PES [35] Tao, F., Qi, Q., Liu, A., Kusiak, A. (2018). Data-driven smart
Transmission and Distribution: Latin America Conference and Exposition manufacturing. Journal of Manufacturing Systems, 48, 157-169.
(T&D-LA) (pp. 1-6). IEEE. [36] Mourtzis, D., Vlachou, E., Milas, N., Dimitrakopoulos, G. (2016). Energy
[13] Guo, W., Zhuang, Z., Yao, J., & Yuan, P. F. (2019, July). Reference consumption estimation for machining processes based on real-time shop
Building Energy Modeling: A Case Study for Green Office Buildings in floor monitoring via wireless sensor networks. Procedia CIRP, 57, 637-
Shanghai. In The International Conference on Computational Design and 642.
Robotic Fabrication (pp. 136-144). Springer, Singapore. [37] Ashok, S. (2006). Peak-load management in steel plants. Applied Energy,
[14] Bachant, P., Wosnik, M., Gunawan, B., & Neary, V. S. (2016). 83(5), 413-424.
Experimental study of a reference model vertical-axis cross-flow turbine. [38] Yu, M., Lu, R., & Hong, S. H. (2016). A real-time decision model for
PloS one, 11(9). industrial load management in a smart grid. Applied energy, 183, 1488-
[15] Alam, K. M., & El Saddik, A. (2017). C2PS: A digital twin architecture 1497.
reference model for the cloud-based cyber-physical systems. IEEE access, [39] Paudyal, S., Cañizares, C. A., Bhattacharya, K. (2014). Optimal operation
5, 2050-2062. of industrial energy hubs in smart grids. IEEE Transactions on Smart Grid,
[16] da Cruz, M. A., Rodrigues, J. J. P., Al-Muhtadi, J., Korotaev, V. V., & de 6(2), 684-694.
Albuquerque, V. H. C. (2018). A reference model for internet of things [40] Geysen, D., et al. (2016). Simplified assessment methodology for optimal
middleware. IEEE Internet of Things Journal, 5(2), 871-883. valorization of flexible industrial electricity demand. EU Project IndustRE,
[17] Ling, J., Feng, Z., Ming, M., & Xiao, X. (2018). Damping controller deliverable D3.3.
design for nanopositioners: A hybrid reference model matching and virtual [41] Wang, J., Huang, Y., Chang, Q., Li, S. (2019). Event-Driven Online
reference feedback tuning approach. International Journal of Precision Machine State Decision for Energy-Efficient Manufacturing System Based
Engineering and Manufacturing, 19(1), 13-22. on Digital Twin Using Max-Plus Algebra. Sustainability, 11(18), 5036.
[18] Lee, K., Webb, S., & Ge, H. (2015). Characterizing and automatically [42] Villar, J., Bessa, R., Matos, M. (2018). Flexibility products and markets:
detecting crowdturfing in Fiverr and Twitter. Social Network Analysis and Literature review. Electric Power Systems Research, 154, 329-340.
Mining, 5(1), 2. [43] O'Connell, N., Madsen, H., Pinson, P., O'Malley, M., Green, T. (2014,
[19] Silva, H. D., Soares, A. L., Bettoni, A., Francesco, A. B., & Albertario, S. October). Regulating power from supermarket refrigeration. In IEEE PES
(2019, September). A Digital Platform Architecture to Support Multi- Innovative Smart Grid Technologies, Europe (pp. 1-6). IEEE.
dimensional Surplus Capacity Sharing. In Working Conference on Virtual [44] Filipe, J., Bessa, R. J., Reis, M., Alves, R., Póvoa, P. (2019). Data-driven
Enterprises (pp. 323-334). Springer, Cham. predictive energy optimization in a wastewater pumping station. Applied
[20] Parker, G. G., Van Alstyne, M. W., & Choudary, S. P. (2016). Platform Energy, 252, 113423.
Revolution: How Networked Markets Are Transforming the Economy? [45] Schulman, J., Wolski, F., Dhariwal, P., Radford, A., Klimov, O. (2017).
and How to Make Them Work for You. WW Norton & Company. Proximal policy optimization algorithms. arXiv preprint
[21] Bauer, R. M., & Gegenhuber, T. (2015). Crowdsourcing: Global search arXiv:1707.06347.
and the twisted roles of consumers and producers. Organization, 22(5), [46] del Campo, G., Calatrava, S., Cañada, G., Olloqui, J., Martinez, R., &
661-681. Santamaria, A. (2018, June). IoT Solution for Energy Optimization in
[22] Benoit, S., Baker, T. L., Bolton, R. N., Gruber, T., & Kandampully, J. Industry 4.0: Issues of a Real-life Implementation. In 2018 Global Internet
(2017). A triadic framework for collaborative consumption (CC): Motives, of Things Summit (GIoTS) (pp. 1-6). IEEE.
activities and resources & capabilities of actors. Journal of Business [47] Verma, N. K., Dev, R., Dhar, N. K., Singh, D. J., & Salour, A. (2017,
Research, 79, 219-227. June). Real-time remote monitoring of an air compressor using MTConnect
[23] Cohen, B., & Kietzmann, J. (2014). Ride on! Mobility business models standard protocol. In 2017 IEEE International Conference on Prognostics
for the sharing economy. Organization & Environment, 27(3), 279-296. and Health Management (ICPHM) (pp. 109-116). IEEE.
[24] Täuscher, K., & Laudien, S. M. (2018). Understanding platform business [48] Tao, F., Zhang, M., Cheng, J., & Qi, Q. (2017). Digital twin workshop: a
models: A mixed methods study of marketplaces. European Management new paradigm for future workshop. Computer Integrated Manufacturing
Journal, 36(3), 319-329. Systems, 23(1), 1-9.
[25] Barnes, S. J., & Mattsson, J. (2016). Building tribal communities in the [49] Zhang, M., Zuo, Y., & Tao, F. (2018, March). Equipment energy
collaborative economy: an innovation framework. Prometheus, 34(2), 95- consumption management in digital twin shop-floor: A framework and
113. potential applications. In 2018 IEEE 15th International Conference on
[26] Manic, M., Amarasinghe, K., Rodriguez-Andina, J. J., & Rieger, C. Networking, Sensing and Control (ICNSC) (pp. 1-5). IEEE.
(2016). Intelligent buildings of the future: Cyberaware, deep learning [50] Karanjkar, N., Joglekar, A., Mohanty, S., Prabhu, V., Raghunath, D., &
powered, and human interacting. IEEE Industrial Electronics Magazine, Sundaresan, R. (2018, November). Digital twin for energy optimization in
10(4), 32-49. an SMT-PCB assembly line. In 2018 IEEE International Conference on
[27] Mathaba, T., & Xia, X. (2015). A parametric energy model for energy Internet of Things and Intelligence System (IOTAIS) (pp. 85-89).
management of long belt conveyors. Energies, 8(12), 13590-13608. [51] Yin, D. R. K. (2009). Case Study Research: Design and Methods. Sage
[28] Chertow, M. R. (2000). Industrial symbiosis: literature and taxonomy. Publications (CA).
Annual review of energy and the environment, 25(1), 313-337.