ArticlePDF Available

Energy management systems by means of computational intelligence algorithms


Abstract and Figures

This work pretends to take advantage of powerful capabilities of computational intelligence to improve the actual features of modeling, prognosis, diagnosis and optimization of load demand for EMS. This work gives a potent complement to the rising new paradigms about renewable energies, distributed generation, micro-grids and smart grids in general, which are in focusing in the optimization or improving of how the energy is generated and not how the energy is used. Peer Reviewed Postprint (published version)
Content may be subject to copyright.
2010 Barcelona Forum on Ph.D. Research in Communications, Electronics and Signal Processing
Energy Management Systems by means of Computational Intelligence Algorithms
Author: J. J. Cárdenas, Thesis Advisor(s): J. L. Romeral, A. Garcia
I. Introduction
In a context of continuous increasing of energy
costs and imperative need of CO2 emission reductions,
the solution is not only look for new clean energy
sources, but the optimization of energy use is also one
of the main ways to solve the problem. We can view the
saved energy as clean energy available for others
consumers. In addition, the saved energy has less
associated costs (economic and human resources).
Hence, the clean energy generation and smart energy
use are two goals to have in mind to get a sustainable
energy production system.
In that way, the interest in Energy Management
(EM) is growing increasingly. Particularly in the field of
industrial and building applications, the Energy
Management Systems (EMS) now are an excellent
option to get this capability and to improve the company
It is in this context that the Information and
Communication Technologies (ICT) take a central role,
and among them, the tools of Artificial Intelligence (AI)
presents an excellent alternative to provide capabilities
of intelligence, autonomy and support in making
decisions for EMS, which has already been done in this
area (and many others as is well known)
This work pretends to take advantage of powerful
capabilities of computational intelligence to improve the
actual features of modeling, prognosis, diagnosis and
optimization of load demand for EMS.
This work gives a potent complement to the rising
new paradigms about renewable energies, distributed
generation, micro-grids and smart grids in general,
which are in focusing in the optimization or improving of
how the energy is generated and not how the energy is
II. Intelligent Energy Management System (iEMS)
In the industrial sector, Energy Management
Systems have focused so far on the monitoring and
“passive” management of energy, as outlined in [1].
Figure 1 presents a diagram that shows the basic
structure of current management systems.
Figure 1. Basic structure of current EMS.
This typical EMS is based on the collection of
information through energy meters (electricity, gas,
water, etc.), which is taken by a SCADA system or
another software for later information management,
whose function is to collect the data, store them and
presented them appropriately for the users. The
software is able also to analyze data and generate
reports to identify critical points of consumption.
The main advantage in these monitoring systems is
to enable the user to control consumption and costs
through energy audits that are supported by data
collected continuously, improving the energy efficiency
of the plant, its processes and devices.
In addition, these energy data could be used to
build up models and get consumption trends. So in
some recent works models based on linear regressions
of energy consumption versus production scheduling
are proposed [2]. These models could be used to
forecast energy consumption for a given production
volume, and this property has a number of useful
applications, ranging from consumption energy
prediction for production scheduling to predicting the
possible energy saving if any proposed energy
efficiency measure is taken.
However, the models used and the mechanisms of
action are simple and not to use advanced analytical
tools for prediction and control, which could improve the
system performance without affect the production or
comfort, and taking into account cost functions which
help to the system to take optimal choices.
Under that context, this thesis proposes an iEMS
(Fig. 2) running advanced algorithms of modeling,
prognosis, optimization and diagnosis.
Figure 2. Block diagram of the proposed iEMS.
It can be seen in Fig. 2 an iEMS that access to
information collected from the meters via field buses. It
also has ability to interact with active systems in the
building or factory plant to take actions with the objective
of optimizing the demanded power.
The iEMS will take advantage of its ability to create
advanced models of load profiles of consumptions from
collected information (energy database) and information
entered by users, which will generate consumptions
models and consumptions prognosis. Based on the
models the iEMS will schedule of power load demand in
order to reduce of power peaks and contracted power.
Moreover, diagnosis and detection of anomalies by
detecting deviations from the models could be carried
In order to get these advanced features, the
algorithms will be based on computational intelligent
tools. For example, optimization algorithms will be
based on evolutionary algorithms, game theory, among
others. Neural networks, fuzzy systems and ANFIS with
support of statistical analysis could be used for
consumption’s models, forecasting, and diagnosis.
III. Smart Use of Energy
Using intelligently the energy according to this work
has three main pillars, which are the expected results of
this work. These are:
III.A. Modeling and prognosis
Building up advanced algorithms of modeling and
prognosis in different levels in an industrial plant or
office building. In the Fig. 3, we can see an example of a
load profile model in function of daily production and
maximum temperature. In the Fig 4, the example of a
power load profile or prognosis.
The modeling part is the base of the proposed
iEMS. If we can know how the load demand will be, we
can take actions to change it.
500 1000 1500 2000
x 10
Producc ión
Tmax C)
Consumo (kWh)
Figure 3. Load profile modeling in function of maximum
temperature and production in an industrial plant.
02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17
Consumpt ion (kWh)
Consumption data vs. date
Real data
Figure 4. Load profile prognosis in an industrial plant.
III.B. Advanced demand management
Implement algorithms of optimization and diagnosis,
taking the former modeling and prognosis algorithms as
base. The Fig. 5 shows an example the demand
optimization where the maximum demanded power has
been reduced by means of a genetic algorithm [3, 4],
which look for the optimum scheduling of consumptions
to avoid peaks and use the same or less energy in the
affected process. The upper graphic in the Fig. 5 is the
load profile without load demand scheduling. The
bottom graphic is the load demand after to apply the
scheduling that has been obtained by the genetic
III.C. Real world Integration
To provide for the EMS a true demand management
and diagnosis system in the field of industrial and
building consumptions. For this reason, it is important
that the proposed algorithms will be easily integrated
0 5 10 15 20 25 30 35
Time (s)
Current (A)
0 5 10 15 20 25 30 35
Time (s)
Current (A)
Total LP
Total LP
Figure 5. Power demand optimization in an automated
with real EMS. In that sense the current algorithms are
been tested in real industrial plants.
IV. Conclusions
This thesis project presents an iEMS framework,
which is blessed with advanced algorithms of modeling,
prognosis, optimization and diagnosis. These
characteristics are aimed to improve the use of energy
in factories and buildings.
The algorithms of computational intelligence are
thought to be one of the best alternatives to implement
the former characteristics.
Some results in modeling and scheduling
optimization have been obtained. The base algorithms
have been the ANFIS for modeling and GA for
optimization. Finally, real data have been used to test
the proposed algorithms in both cases, getting the
expected results.
V. Acknowledgments
Part of this thesis work is been supported by
projects of technology transfer, particularly we can
acknowledge the support of SEAT that has permit us the
use of their energy database, under the project named
VI. References
[1] J. Sheppard and A. Tisot, "Industrial Energy Management:
Doing More with Less," IETC - Industrial Energy
Technology Conference, 2006.
[2] J. C. Van Gorp, "Enterprising energy management," Power
and Energy Magazine, IEEE, vol. 2, pp. 59-63, 2004.
[3] J. Cardenas, A. Garcia, J. L. Romeral, and J. Urresty, "A
Multi-Objective GA to Demand-side Management in an
Automated Warehouse," in ETFA2009, the IEEE 2009
Conference on Emerging Technologies and Factory
Automation, Palma de Mallorca, Spain, 2009.
[4] J. J. Cardenas, A. Garcia, J. L. Romeral, and F. Andrade,
"A genetic algorithm approach to optimization of power
peaks in an automated warehouse," in Industrial
Electronics, 2009. IECON '09. 35th Annual Conference of
IEEE, 2009, pp. 3297-3302.
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
The simultaneous operation of the automated storage and retrieval machines (ASRs) in an automated warehouse can increase the likelihood that high power demand peaks tum unstable the electric system. Furthermore, high power peaks mean the need for more electrical power contracted, which in turns leads to more fixed operation cost and inefficient use of the electrical installations. In this context, we present a multi-objective genetic algorithm approach (MOGA) to implement demand-side management (DSM) in an automated warehouse. It works minimizing the total energy demand, but without increasing substantially the time for the operation. Simulations show the performances of the new approach.
The adaptation of enterprise energy management (EEM) technology by businesses to control costs, optimize processes, and prevent downtime, is discussed. The system can provide facility managers with the information to make informed decisions, from a functional perspective and a financial one. It consists of a network of intelligent energy meters linked to a centrally located server running the enterprise management software. The system logs the information in an historical database, responds to any alarm conditions by relaying notifications to operations personnel, and displays the real-time statues of each monitored area on the screens of one or more networked workstations.
Over the last several decades, there has been increasing interest and activity in the filed of energy management. The knowledge gained from the deployment of energy efficient projects is driving a transition from traditional tactical practices to more comprehensive strategic energy management practices. Like many modern management practices, a strategic management approach highlights the need for an information system to set goals, track performance, and communicate results. An enterprise energy management (EEM) information system can support strategic energy management by capturing data about current energy performance, assisting in the creation and tracking of energy performance goals, and communicating results to managers and energy project participants. This paper presents specific EEM system applications that support strategic energy management, which include: modeling and forecasting; benchmarking; energy use and cost analysis; and measurement and verification.