Laboratório de Sistemas Eléctricos Industriais
|
|
Laboratório de Sistemas Eléctricos Industriais |
|
Papers in International Journals with Referee or Books
|
|
Communications in scientific meetings with Referee
|
|
Technical Reports
|
| Automatic Language Control of Electrical Drives. Background and Applications |
A new formal language concept for
automatic control of electrical drives is presented. The proliferation
of these systems promises improved efficiency and reliability, reduced
maintenance and operative costs, and a more environment friendly
operation. However, they exhibit considerable complexities related with
their behaviour such as large uncertainties at structural and parameter
levels, multidimensionality, and strong mutual interactions. The
integration of these systems with information processing techniques
allows a better adaptative modelling of their behaviour. The electrical
drive system is assumed as a linguistic source producing a certain
language. A developed grammatical inference algorithm is used in order
to extract the productions that govern a grammar representing that
language. The formalism of the developed formal language based control
algorithm is presented and experimental results are discussed. |
| Supervision Language Control of Electromechanical Drives |
The aim of this paper is to introduce a formal
language based approach to the automatic supervision control of
electromechanical drives. Modern applications often involve
electromechanical drives with high dynamical complexity. The
proliferation of these systems promises improved efficiency and
reliability, reduced maintenance and operative costs, and a more
environment friendly operation. However, they exhibit considerable
complexities related with their behavior such as large uncertainties at
a structural and parameter levels, multidimensionality, and strong
mutual interactions. The integration of these systems with information
processing techniques allows a better adaptative modeling of their
behavior and thus better supervision concepts. Formal language
techniques have been used in the past to study autonomous dynamical
systems. Assuming the drive system as a linguistic source, producing a
certain language, the modeling framework is based on a formulation in
terms of context-dependent grammars, distinguishing between information
generated by the system and by the input control. Making use of the
formal language descriptive characteristics the grammatical supervision
algorithm governs the overall behavior of the drive. The formalism of
the formal language based supervision control algorithm is presented and
experimental results are discussed. |
| SVM Voltage Control implementation using low-cost reconfigurable FPGA system |
This paper describes a full VHDL implementation of a
voltage Space-Vector Modulation in a FPGA reconfigurable system. A VHDL
code was specially developed and optimized based on a finite state
machine, which has been proven a more flexible and integrated
environment for performing the SVM. A FPGA XC4005XL from Xilinx is used
to allow a high economical solution. MATLAB functions automatically
optimize and generate a configuration bitstream for the FPGA.
Experimental results confirm the validity of the proposed methodology. |
| Design and Implementation of a Laboratory Pseudo‑Master Oscilloscope Network |
This paper presents the design, development and
implementation of a laboratory pseudo-master oscilloscope network. The
purpose of the network is to centralize the data from a series of
several laboratory working benches into only one regular PC. The
decision to store the experimental data is taken in each working bench,
running the oscilloscopes as pseudo-master units. In order to cover
large inter-benches distances the RS-485 protocol was adopted to
establish the network over several oscilloscopes. A management software
was developed to manage the data flow through all the network, the
storage, in different folders, and visualization of the data obtain from
the oscilloscopes. The stored data can be sent directly to a printer or
embedded in other software applications. The potential of the developed
system is shown, namely regarding to flexibility, cost and performance,
in the scope of laboratory working quality. |
| Formal Language Modelling of a Switched Relutance Machine |
The Switched Reluctance Machine is being widely
spread among industrial and domestic applications, replacing other
electrical machines. However its modelling and control present important
difficulties, due to complex system nonlinearities. In this paper it is
shown how the use of formal language technique can help in the modelling
of a Switched Reluctance Machine, overcoming these difficulties.
Experimental tests with a lab prototype are performed in order to obtain
a realistic and automatic machine model. |
| A Network Distribution Power System Fault Location Based on Neural Eigenvalue Algorithm |
A new approach to fault location for distribution
network power system is presented. This approach uses the Eigenvalue and
an artificial neural network based learning algorithm. The neural
network is trained to map the non-linear relationship existing between
fault location and characteristic Eigenvalue. The proposed approach is
able to identify, to classify and to locate different types of faults
such as: single-line-to-ground, double-line-to-ground, double-line and
three-phase. Using the Eigenvalue as neural network inputs the proposed
algorithm is able to locate the fault distance. The results presented
show the effectiveness of the proposed algorithm for correct fault
diagnosis and fault location on a distribution power system networks. |
| A neural space vector fault location for parallel double-circuit distribution lines |
A new approach to fault location for parallel
double-circuit distribution power lines is presented. This approach uses
the Clarke–Concordia transformation and an artificial neural network
based learning algorithm. The α, β, 0 components of double
line currents resulting from the Clarke–Concordia transformation are
used to characterize different states of the system. The neural network
is trained to map the non-linear relationship existing between fault
location and characteristic eigenvalue. The proposed approach is able to
identify and to locate different types of faults such as:
phase-to-earth, phase-to-phase, two-phase-to-earth and three-phase.
Using the eigenvalue as neural network inputs the proposed algorithm
locates the fault distance. Results are presented which shows the
effectiveness of the proposed algorithm for a correct fault location on
a parallel double-circuit distribution line. |
| Implementation of an On-Line Learning Speed Controller for a Switched Reluctance Machine |
In the classical speed control of electrical machines
it is usual to have linear PID controllers. This kind of controllers can
present good results near a given operating point, where the machine may
be considered as a linear system. However, the classical PID controller
may have some limitations if it is planed to be used in a wide working
region. In this case a good alternative can be the learning controllers
using fuzzy logic and neural networks.In this paper it is presented and
mainly discussed a neuro-fuzzy learning controller for speed regulation
of an 8/6 Switched Reluctance Machine. Experimental results are
presented and discussed showing how the controller can learn in real
time the “good” rules. The results obtained with a classical PID
controller are shown as a reference speed controller for the analysis.
|
| Performance of a Four Phase Switched Reluctance Motor Speed Control Based On an Adaptive Fuzzy System: Experimental Tests, Analysis and Conclusions |
In this paper,
the controller’s tuning and performance of a speed controller prototype
for switched reluctance motor is presented by a number of significative
experimental tests. The system uses an on-line learning mechanism to
acquire and modify, if needed, the “good” fuzzy control rules.
Experimental essays are analysed and discussed to reveal some advantages
of having a learning speed controller to the SR machine, and also the
drawbacks that using these controllers can introduce to the drive system
and possible ways to overcome them. |
| Metodologia de Parametrização de um Controlador Neuro-Fuzzy de Velocidade para uma Máquina de Relutância Variável | In the classical speed control of electrical machines it is usual to
have linear PID controllers. This kind of controllers can present good
results near a given operating point, where the machine can be
considered as a linear system. However, the classical PID controller has
some limitations if it is planned to be used in a wide working region of
electrical machines. In this case a good alternative can be the adaptive
controllers using fuzzy logic and neural networks. One problem is
tunning all parameters of the controller. This work presents the applied
method and experimental results. |
| Implementation of a Neuro-Fuzzy Speed Controller for a Switched Reluctance Machine | In the classical speed
control of electrical machines it is usual to have linear PID
controllers. This kind of controllers can present good results near a
given operating point, where the machine can be considered as a linear
system. However, the classical PID controller has some limitations if it
is planed to be used in a wide working region of electrical machines. In
this case a good alternative can be the “intelligent” controllers using
fuzzy logic and neural networks. |
| Obtaining the Magnetic Characteristics of an 8/6-Switched Reluctance Machine: FEM Analysis and Experimental Tests |
This paper describes the step-by-step procedure for modeling an 8/6
switched reluctance machine. Starting from establishing the geometry of
the machine, a nite element modeling approach is initially developed to
compute the [1]ux linkage/current/rotor position relationship extracting
proper and mutual inductances, also its torque and eld distribution
characteristics. Experimental tests are performed to obtain the
characteristics. Experimental tests are performed to obtain the magnetic
characteristics of themachine, comparing, correcting, and discussing the
results obtained with those of FEM analysis. |
|
This paper describes the systematic procedure for designing a MOSFET
power converter for an 8/6-switched reluctance machine. Starting from
establishing its topology and subsystems, a Matlab simulation approach
is initially developed to design its driving circuits, test different
converter operating modes, and phase current controllers. Experimental
tests are performed to verify the converter functionality and
performance, comparing, correcting and discussing the results obtained
with simulation analysis. |
|
| On the use of Reactive Power as an Endogenous Variable in Short-term Load Forecasting |
In the last decades, short-term load forecasting (STLF)
has been the object of particular attention in the power systems field.
STLF has been applied almost exclusively to the generation sector, based
on variables, which are transversal to most models. Among the most
significant variables we can find load, expressed as active power (MW),
as well as exogenous variables, such as weather and economy related ones,
although the latter are applied in larger forecasting horizons than STLF. |
| Estimating load diagrams in electricity substations | Load forecast is dealt with in this paper from the point of view of
distribution networks. The issues associated with network management, as
load dispatch and network reconfiguration, under quality of service
constraints, require reliable short-term (next hour) load forecasts.
Maintenance issues and eventual power purchase decisions within
liberalised electricity markets require, among others, reliable next-day
load forecasts. Substations in electricity distribution networks are
usually monitored to such an extent that power and energy data is
available at dispatch centres with a reasonable spatial and time detail,
as provided by SCADA systems spread throughout the networks. This
availability allows forecast models to be tailored to specific networks,
in such a way that it is possible to grasp the particular load behaviour
and provide forecasts with a good level of confidence. The paper deals
with a methodological approach to the use of artificial neural networks
(ANN) as forecast models. It makes use of a procedural sequence for the
pre-processing phase that allows capturing certain predominant relations
among certain different sets of available data, providing a more solid
basis to decisions regarding the composition of the input vector to ANN.
The methodological approach is discussed and a real life case study is
used for illustrating the defined steps, the ANN and the quality level
of the results. |
| Load Forecast Using Trend Information ANN Process Reconstruction |
The algorithms for short-term load forecast (STLF),
especially within the next-hour horizon, belong to a group of
methodologies that aim to render more effective the actions of planning,
operating and controlling electric energy systems (EES). In the context
of the progressive liberalisation of the electricity sector, unbundling
of the previous monopolistic structure emphasizes the need for load
forecast, particularly at the network level. Methodologies such as
artificial neural networks (ANN) have been widely used in next-hour load
forecast. Designing an ANN requires, amongst other things, the proper
choice of input variables, avoiding overfitting and an unnecessarily
complex input vector (IV). This may be achieved by trying to reduce the
arbitrariness in the choice of endogenous variables. At a first stage,
we have applied the mathematical techniques of process-reconstruction to
the underlying stochastic process, using coding and block entropies to
characterize the measure and memory range. At a second stage, the
concept of consumption trend in homologous days of previous weeks has
been used. The possibility to include weather-related variables in the
IV has also been analyzed, the option finally being to establish a model
of the non-weather sensitive type. The paper uses a real-life case
study. |
| A New Control Strategy Based On Optimised Smooth-Torque Current Waveforms for Switched Reluctance Motors |
With the purpose of improving the dynamic performance
of the Switched Reluctance Motor (SRM) one analyses its operation using
MMF waveforms imposed in the linear region of its magnetic
characteristics. Three factors are important in the control strategy:
the dwell angle concerning the phase current and its location in the
phase inductance profile; the phase current profiling; and the current
control suitable to the application of the adopted strategy. In this
paper, the first two points are analysed and a methodology is proposed
to reduce the torque ripple and attenuate the phase voltage, by
modelling current waveforms with specific criteria. |
| Network Operating Characteristics Based on Imposed MMF Waveforms for Switched Reluctance Generators |
With the propose of improving the dynamic performance
of the Switched Reluctance Generator (SRG) one analyses its operation
using magnetomotive force (MMF) waveforms imposed in the linear region
of the magnetic characteristics. In this paper, based on the proposed
location of the dwell angle in the phase inductance profile, supported
by energetic criteria, the SRG network operating characteristics were
analysed for two typical cases: feeding a passive (ohmic) load and
connected to a power network (infinite busbar). |
| A Methodology Based on Energy-Conversion Diagrams to Improve Switched Reluctance Generators Control | With the aim of improving the dynamic performance of the Switched
Reluctance Generator one analyses its operation using magnetomotive
force (MMF) waveforms imposed in the linear region of its magnetic
characteristics. Three factors are important in the control strategy of
a SRG: the dwell angle concerning the phase current and its location in
the inductance profile; the phase current profile; and the current
control. In this paper, the first point is analyzed and a methodology
based on energetic aspects is proposed in order to achieve higher
efficiency. |
| Entropy Analysis on Electrical Drives Language Modeling | This paper discusses the relationship between the memory range of a
controlled dynamical system and the maximum type of allowed productions
in a grammatical learning algorithm. This is an important issue when the
performance of the learning process is discussed. The shorter the
maximum type of allowed productions is the faster the learning process
will be. However small values assigned for the maximum type of allowed
productions can damage the learning process, being the obtained grammar
an insufficient model of the electrical drive. The memory range of a
dynamical system can be achieved by using mathematical techniques of
process-reconstruction, using coding and block entropies. These modeling
techniques allow a precise knowledge of the electrical drive dynamics,
fundamental topic in modern control approaches. |
| The Switched Reluctance Generator for Wind Power Conversion | The use of wind energy has become increasingly important as a
renewable energy source, and therefore there is an increasing interest
in exploiting it using a Switched Reluctance Machine (SRM) as a
generator and optimise its characteristics in this domain. It is in this
context that this paper analyzes the generator mode of the SRM 1) in a
direct coupling to the turbine shaft and 2) coupled to the shaft through
a gearbox. This paper is aimed at analyzing and proposing an alternative
technical solution for wind power conversion, regarding the electrical
generator system. |
| O Gerador Eléctrico de Relutância Comutada como Alternativa aos Geradores Clássicos nos Aproveitamentos de Energia Eólica | A redução de emissões gasosas nefastas para o meio ambiente e a
diminuição da dependência do petróleo como fonte primária de energia,
são as principais razões que conduzem ao interesse crescente nas
energias renováveis. De entre os vários tipos de energias renováveis
existentes, a eólica é onde recai maior atenção, reflectindo-se
naturalmente na investigação e desenvolvimento, particularmente nos
sistemas electromecânicos e seus componentes, entre os quais os
geradores eléctricos. Actualmente, as máquinas síncrona e assíncrona
dominam o panorama das aplicações de energia eólica, no entanto, a
Máquina Eléctrica de Relutância Comutada (MERC) tem sido alvo de
investigação e é apontada como alternativa válida nestes aproveitamentos
de energia. A despeito da MERC estar mais divulgada como motor, quer em
aplicações de carácter industrial quer doméstico, como gerador existem
aplicações na indústria aeronáutica e ainda, em menor número, trabalhos
no domínio de geradores em aproveitamentos de energia eólica . |
| A Simple PID Controller with Adaptive Parameter in a dsPIC; Case of Study | The main goal of this work consists in the development and
implementation of a discrete PID controller with fast response and
parameters adaptation capability, in an automatic way. This controller
is based on a classic PID where a parameters adaptation algorithm was
associated in order to control a process. This PID do not require any
kind of adjustment or calibration from the operator. For the parameters
adaptation one fuzzy system with a Takagi-Sugeno inference mechanism was
chosen and some simplification of this system of decreasing the
processing time and the controller response (250ms), in order to control
fast processes without losing stability. The developed algorithm was
implemented in a recent dsPIC30F.
|
| An Adaptive Learning Rate Approach for an On-Line Neuro-Fuzzy Speed Controller Applied to a Switched Reluctance Machine | The mostly used neuro-fuzzy motor speed control systems are time
consuming and have an high computation effort when the speed reference
changes gradually and the system has to learn the new operating point
most of the time. In this cases a degradation of the system performance
is evident has is demonstrated by experimental results in this paper. To
surpass these effects, a decision and adaptation algorithm of the
learning rate applied to the neuro-fuzzy control’s approach is proposed.
The adaptive learning rate algorithm with the controller is tested and
compared in the speed control system for an 8/6 switched reluctance
motor by experimental tests. The proposed solution is explained, tested
and the experimental tests are presented and discussed.
|
| A Neuro-Fuzzy Multilayer Weights Approach for an On-Line Learning Speed Controller Applied to a Switched Reluctance Machine: Why and How to Use | The most used neuro-fuzzy motor speed control systems are time
consuming and have an high computation effort when the speed reference
changes suddenly and the system, most of the time, has to learn this new
operating point. In this case a degradation of the system performance is
evident, as is demonstrated by our experimental results in this paper.
To surpass these effects, a new neuro-fuzzy multilayer control’s
approach is proposed. The multilayer controller is tested and compared
in the speed control system for an 8/6 switched reluctance motor by
experimental tests. The proposed solution is explained, tested and the
experimental results are presented and discussed.
|
| On-Line Diagnosis of Three-Phase Induction Motor Using an Eigenvalue alfa beta–Vector Approach | An on-line diagnosis of three-phase induction motor using an
automatic algorithm based on Park’s vector approach is presented. This
algorithm uses the Eigenvalue approach and allows an interpretation of
the ab–vector. In fact, stator current ab–vector patterns are first
obtained, then the Eigenvalue algorithm is used to see if the motor is
healthy or not. Experimental results are presented in order to show the
effectiveness of the proposed method.
|
| On the Use of Matlab/Simulink as a Tool for the Study of Power Systems Transient | This paper presents a methodology for the modelling of power systems
on the transient analysis. In this methodology the MATLAB/SIMULINK"
program is used. This program width their non linear blocks can be
advantageously used for different kind of variable situations on
electrical power networks, such as fault diagnosis and stability
control. Furthermore, the modelling effort and required simulation times
are smaller than the conventional simulators, such as EMTP. Since state
space models are used, this simulation tool is useful to study the
transient phenomena and easier to compare with power systems laboratory
results. This paper summarizes the used method and gives some examples.
|
| A Neural Space Vector Fault Location for Parallel Double-Circuit Distribution Lines | A new approach to fault location for parallel double-circuit
distribution power lines is presented. This approach uses the
Clarke-Concordia transformation and an artificial neural network based
learning algorithm. The a,b,0 components of double line currents
resulting from the Clarke-Concordia transformation are used to
characterize different states of the system. The neural network is
trained to map the non-linear relationship existing between fault
location and characteristic Eigenvalue. The proposed approach is able to
identify and to locate different types of faults such as:
Phase-to-Earth, Phase-to-Phase, Two-Phase-to-Earth and Three-Phase.
Using the Eigenvalue as neural network inputs the proposed algorithm
locates the fault distance. Results are presented which shows the
effectiveness of the proposed algorithm for a correct fault location on
a parallel double-circuit distribution line.
|
| An Average Values Global Model for the Switched Reluctance Machine | The main subject of this paper is to present a new simplified global
model for the Switched Reluctance Machine (SRM), useful, namely, for the
command and control analysis of this kind of system. The concepts of
power and energy are used in the global model construction, with the
definition of global parameters and variables. This new global model
presents an advantage over the classical detailed one, which is the
reduction of the number of dynamic equations. The time-dependent global
parameters disadvantage, which appears in the model construction
process, is overcome with the consideration of average values for global
variables and global parameters, allowing the representation of the
machine non-detailed behavior. An Average Values Global Model is
obtained. An 8/6 SRM with four phases, eight stator poles and six rotor
poles is characterized and used for illustration of the system behavior,
regarding its variables evolution. In the SRM command it is assumed that
the m.m.f. is imposed, which means that the system current is
controlled.
|
| A Model for the Switched Reluctance Machine with Global Parameters and Global Variables | O objectivo principal deste artigo é apresentar um modelo para a
Máquina Eléctrica de Relutância Comutada, útil na análise do comando e
controlo deste tipo de máquina. Utilizando os conceitos de potência e
energia construiu-se um modelo com variáveis e parâmetros globais, que
serão definidos no artigo. Este modelo tem a vantagem formal sobre o
modelo detalhado tradicional de ter um reduzido número de equações
dinâmicas. Esta abordagem pode ser vista como um caminho inicial no
sentido de uma maior simplificação do modelo da máquina recorrendo a
aproximações na determinação dos parâmetros e variáveis. Para ilustrar o
comportamento do sistema, através da evolução dos vários parâmetros e
variáveis, é utilizada uma máquina 8/6, com quarto fases, oito pólos no
estator e seis pólos no rotor. No comando da máquina será assumido que a
força magnetomotriz é imposta, o que significa que a corrente do sistema
será controlada. Isto é útil na optimização do comportamento da máquina.
|
| Fixed-Frequency Active Current Controller and Low-Sensitivity Voltage Regulator for a Voltage Sourced Buck-Boost Type Rectifier | We present an active fixed-frequency input current controller and a
low-sensitivity output voltage regulator for a voltage-source buck-boost
type rectifier operating in the continuous current mode. The current
controller enforces the input line current to track the desired
sinusoidal waveform in phase with the input source voltage, ensuring a
near-unity power-factor operation of the rectifier. The current control
strategy actively shapes the input current allowing a near sinusoidal
input current even with high variations in the DC inductor current. A
low-sensitivity output voltage regulator modulating the amplitude of the
sinusoidal reference for the current controller is also proposed. Both
the current controller and the voltage regulator are based on
sliding-mode theory. Experimental results are presented and the
performance compared to that obtained using a conventional Proportional
Integral (PI) output voltage controller.
|
| A Criteria for Designing Switched Reluctance Motors with Torque Ripple Reduction | With the goal of improving the dynamic performance of a Switched
Reluctance Motor (SRM), it is important to find control strategies that
reduce the machine torque ripple. In this way the settlement of a
reference MMF synchronised with the rotor position, in conjunction with
a suitable design of the SRM, allow to minimize the torque ripple,
maintaining the torque nominal level of this type of machine. In this
paper, an attention is devoted to the SRM design, concerning especially
the relation between the number of stator and rotor poles. It is defined
a criteria, based in the construction parameters of the machine, that
indicates, a priori, which are the machine configurations that allow the
minimization of torque ripple.
|
| Short-term load forecast using trend information and process reconstruction | The algorithms for short-term load forecast (STLF), especially
within the next-hour horizon, belong to a group of methodologies that
aim to render more effective the actions of planning, operating and
controlling electric energy systems (EES). In the context of the
progressive liberalization of the electricity sector, unbundling of the
previous monopolistic structure emphasizes the need for load forecast,
particularly at the network level. Methodologies such as artificial
neural networks (ANN) have been widely used in next-hour load forecast.
Designing an ANN requires the proper choice of input variables, avoiding
overfitting and an unnecessarily complex input vector (IV). This may be
achieved by trying to reduce the arbitrariness in the choice of
endogenous variables. At a first stage, we have applied the mathematical
techniques of process-reconstruction to the underlying stochastic
process, using coding and block entropies to characterize the measure
and memory range. At a second stage, the concept of consumption trend in
homologous days of previous weeks has been used. The possibility to
include weather-related variables in the IV has also been analysed, the
option finally being to establish a model of the non-weather sensitive
type. The paper uses a real-life case study. |
| Unsupervised Neural-Network Based Algorithm for an On-Line Diagnosis of Three-Phase Induction Motor Stator Fault |
In this paper an automatic algorithm based an unsupervised
neural-network for an on‑line diagnostics of three-phase induction motor
stator fault is presented. This algorithm uses the alfa‑beta stator
currents as input variables. Then a fully automatic unsupervised method
is applied in which a hebbian-based unsupervised neural-network is used
to extract the principal components of the stator current data. These
main directions are used to decide where the fault occurs and a
relationship between the current components is calculated to verify the
severity of the fault. One of the characteristics of this method, given
its unsupervised nature, is that it does not need a prior identification
of the system. The proposed methodology has been experimentally tested
on a 1kW induction motor. The obtained experimental results show the
effectiveness of the proposed method. |
| Entropy Based Choice of a Neural Network Drive Model | The design of a
neural network requires, among other things, a proper choice of input
variables, avoiding over fitting and an unnecessarily complex input
vector. This may be achieved by trying to reduce the arbitrariness in
the choice of the input layer. This paper discusses the relation between the memory range of a particular controlled dynamical system (induction drive) and the dimension of the neural network input vector. Mathematical techniques of process-reconstruction of the underlying process, using coding and block entropies to characterize the measure and memory range, were applied. These modeling techniques provide a precise knowledge of the drive dynamics, a fundamental requirement in modern control approaches. |
| WPEC - A New Web Tool for the Power Electronics Learning | This paper presents a new web tool for the power electronics
learning. This tool has several goals. The first goal is to teach power
electronics through interactive animations. To support the student’s
learning in a more active way a chat option is also provided. This
interactive option allows the students to communicate with the teacher
and with other students. In order to test and verify the students
knowledge an option with questions are also provided. This option
provides several oriented questions in order to identify some problems
in the learning of the students. This last tool will give to the
students a path where they should focus their study. |
| Short-term load forecasting based on ANN applied to electrical distribution Substations | The short-term load forecasting (STLF) algorithms belong to the set
of methodologies which aim to furnish more effectiveness in planning,
operation and conduction in electric energy systems (EES). The presence
of a de-regulated environment reinforces the need of forecast,
particularly in distribution networks. Actions like network management,
load dispatch and network reconfiguration, under quality of service
constraints, require reliable short-term (next hour) load forecasts.
Artificial neural networks (ANN) are widely used in this horizon of
prevision, with satisfactory results. The construction of an “efficient”
ANN goes through, among, other factors, the construction of an
“efficient” input vector, in order to avoid over fitting problems and
keeping the global simplicity of the model. This paper deals with a
methodological approach, in order to provide more solid basis decision
regarding the composition of the input vector, namely, in the choice of
the number of the contiguous values of the principle variable (active
power). In a first approach it was established a search for any “chains
with complete connections”, in the active power signal, based on Gibbs
measure, and a relative entropy analysis. It was introduced the concept
of “consumption tendency” in past homologous days. It was also analyzed
the correlation between the consumption and the climatic data, having
been established a non-weather sensitive model. The methodological
approach is discussed and compared with another input vector. The model
was tested in a real life case study for illustration of defined steps. |