International Review of

Electrical Engineering
(IREE)

PART

Contents

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved

(continued from Part A)

Probabilistic LV Distribution Network Design for Aggregated Light Industrial Loads
by P. van Rhyn, J. H. C. Pretorius, R. Herman

3435

Multi-Objective Environmental Optimal Power Flow Considering Transient Stability

Improvement
by Ramin Moslemi, Masoud Esmaili, Heidar Ali Shayanfar, Lingfeng Wang, Robert C. Green

3443

Artificial Intelligence Forecast of PV Plant Production for Integration in Smart Energy Systems
by Mikhail Simonov, Marco Mussetta, Francesco Grimaccia, Sonia Leva, Riccardo E. Zich

3454

A Novel Approach for Optimal Determination of Number of Distributed Generation Units

Along with their Sizes and Locations
by M. R. Alemi, M. Fallahnejad, M. Sedighizadeh

3461

Linear Programming for Optimal Sizing of DGs in MicroGrid Considering Loss of Power Supply
Probability Technology
by H. A. Zamani, S. M. Ezzati, M. D. Farashah, E. Dahri, S. M. M. Tafreshi

3470

Probabilistic Approach to Reliability Assessment of Electric Power System Containing

Distributed Generation
by Shengyou Xu, Minyou Chen, Li Ran

3478

Solution for Multi-Objective Reactive Power Optimization Problem Using Fuzzified

Particle Swarm Optimization Algorithm
by D. Silas Stephen, M. Devesh Raj, P. Somasundaram

3486

Stabilization of Multi-Machine Power Systems Using Craziness Based Particle Swarm

Optimization and Harmony Search Algorithms
by M. R. Banaei, S. Abbasi, E. Salary

3495

Lightning Transient Modeling of Wind Turbine Towers

by Xiaoqing Zhang, Chenhua Liu

3505

Power System Stability Enhancement Using SSSC

by M. R. Banaei, E. Salary

3512

Precise Modeling and Detection of High Impedance Faults in Microgrid System Based on
Residual Current Harmonic Analysis
by L. Eslami, R. Keivanian

3523

Single Phase Fault Distance Estimation in Medium Voltage Distribution Network Based on
Traveling Waves
by eila Gruhonji Ferhatbegovi, Ante Marui, Ivica Pavi

3532

Blended Crossover Continuous Ant Colony Optimization and Stability Index Tracing for
Effective FACTS Devices Installation
by Z. Hamid, I. Musirin, M. N. A. Rahim

3542

Using Clonal Selection Algorithm for Optimal Placement and Sizing of Unified Power Flow
Controllers in Deregulated Power Systems
by A. A. Ramezani, Y. Ramezani, A. A. Ramezani, M. Ramezani

3554

(continued)

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved

A Detailed Comparison of Cumulant-Based Probabilistic Power Flow Methods

by Duong D. Le, C. Bovo, A. Berizzi, E. Ciapessoni, D. Cirio, A. Pitto

3562

Adaptive Least Squares Support Vector Machines Method for Short-Term Load Forecasting
Based on Mutual Information for Inputs Selection
by Milo B. Stojanovi, Milo M. Boi, Milena M. Stankovi, Zoran P. Staji

3574

A Fault Direction Discrimination Scheme Based on Transients Generated by Faults in MultiBranch Power Systems
by S. Milad Tayebi, Ahad Kazemi, Roshanak Rezaeipour

3586

A Honey Bee Foraging Approach to Optimal Capacitor Placement with Harmonic Distortion
Consideration
by M. Sedighizadeh, F. Kalimdast

3592

Investigation of the Methods for Single Area and Multi Area Optimization of a Power System
Dispatch Problem
by Senthil Krishnamurthy, Raynitchka Tzoneva

3600

A Study on Scheduling of Primary Frequency Control Reserve

by M. Rouholamini, M. Rashidinejad

3628

BDA Fault Detection in Complex Electric Power Systems

by Yagang Zhang, Zengping Wang, Shuqiang Zhao

3638

Static and Dynamic Security-Augmented Market-Based Generation Scheduling Using

Multi-Objective Artificial Bee Colony Algorithm
by M. Gitizadeh, J. Aghaei

3646

An Efficient Non-Convex Power Dispatch Approach for Modern Thermal Units

by Po-Hung Chen, Deng-Fa Lin, Cheng-Chien Kuo

3657

Environmental/Economic Power Dispatch Using Multi-Objective Honey Bee Mating

Optimization
by Javad Javidan, Ali Ghasemi

3667

International Review of Electrical Engineering (I.R.E.E.), Vol. 7, N. 1

ISSN 1827- 6660
January-February 2012

Artificial Intelligence Forecast of PV Plant Production

for Integration in Smart Energy Systems
Mikhail Simonov1,2, Marco Mussetta1, Francesco Grimaccia1, Sonia Leva1, Riccardo E. Zich1

Abstract Computation techniques play an important role in most engineering problems in

which optimization problems have to be faced. Energy management operations represent one of
these cases where real-time energy production, transfer, storage and consumption need to be
optimized. In this context renewable energy sources can be managed using evolutionary
computation and other tools. In this light artificial neural network solution based on weather
forecast can estimate energy flows combined with the event-driven variability encouraging
photovoltaic integration with the electric power system.
This article discusses the role of these computational tools and some issues related to the
variability and uncertainty in the operations where PV plants are potentially fully connected to the
power grid in a future scenario. Copyright 2012 Praise Worthy Prize S.r.l. - All rights
reserved.
Keywords: PV Forecasting, Renewable Energy, Smart Grids, Artificial Intelligence

I.

Introduction

It is well known that variability and uncertainty are

inherent characteristics of power systems concerning
loads, power lines, and generator availability.
Regulations and procedures are conducted to guarantee a
certain level of performance and power quality, based
essentially on forecasting, scheduling, dispatching, and
reserves in the most economic way [1].
In the current energy production state-of-the-art, a
continuous growing contribution comes from
photovoltaic (PV) plants and other renewable sources
which operate independently. However in a scenario with
a massive use of solar electricity generation technologies
their cooperative behavior is still limited to the
simultaneous injection of the energy flows in the grid.
Common production plants often are committed to
generate electricity for the system in a time scale which
comes from hours to several days. Thus operators are
able to change system parameters to follow the load.
More accuracy in the prediction is needed to ensure
that errors in forecasts do not compromise reliability of
the system. On the load side as well forecasting
represents one of the main issues especially for small
power systems which have vital value for electricity
market players like retailers and independent system
operators [2]. Moreover to reach a proper demand-supply
matching it is important to consider that the load profile
used by operators can be sensibly different from the
correspondent individual customer, especially where
some PV production units are present in a specific
residential area.

Manuscript received and revised January 2012, accepted February 2012

3454

Independent business operators tend to run production

sites to maximize their individual objective functions in
terms of energy and economic efficiency, with no other
cooperation forms. They can take autonomously
decisions about PV maintenance and estimation of
energy quantities using local information related to local
conditions and local production dynamics, without any
additional knowledge about the happening in the
neighborhood.
The load forecast in such a condition is probabilistic,
requiring no networking with other PV plants: their local
bus systems remain isolated and not integrated in a wider
network. The renewable PV energy is less-predictable
because it depends on weather conditions, which might
limit the maximum percentage of renewable energy
being injected in the grid because of power quality
issues. Recently different approaches have been used for
PV forecasting based on both physical and statistical
methods [3], but many studies are mainly focused on
short term load forecasting in energy management
systems (EMS) [4][8] since the intrinsic
tight
correlation of scheduling, planning and operations of
power systems.
Forecasting models based essentially on artificial
neural networks (ANN) has been widely applied in such
systems because of their high forecasting accuracy and
self-learning ability, and forecasting accuracy
significantly impacts the cost of all power utilities in the
planning of energy supply [9], [10]. In this context the
authors propose an integrated artificial intelligent
prediction tool applied on the production side which can
be used to forecast variable energy generation sources in
order to allow system operators to plan actions in
Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved

Mikhail Simonov, Marco Mussetta, Francesco Grimaccia, Sonia Leva, Riccardo E. Zich

advance and thus to let increase the develop and

diffusion of renewable generation technologies into
smart energy systems [11], [12].
Among the general issues considered when operating
power systems with variable generation, power quality
(e.g. voltage flicker) has a time dimension in terms of
seconds. However, it is also important to characterize
variability along a spatial dimension that, for power
quality issues, may be in the range of tens of square
kilometers [13].

II.

1000

800

600

400

200

Forecasting PV production

Power [kW]
Irrad [W/m2]

Different methods can be used to forecast solar

resources at differing time scales. Forecasts of PV output
are required for days ahead down to hours ahead and
they should provide information about the expected
output and the degree of uncertainty to predict particular
volatile periods.
The rationale for predictability importance is justified
by the fact that the addition of variable generation
source, like PV or wind farm for example, reduce system
stability that must be managed accordingly. In fact,
general important aspects to consider in power systems
with variable generation sources at different time scales
regard for example, from seconds to days, power quality
issues, regulation reserves, load following and
scheduling actions.
Problems with power quality are often managed
within a single distribution feeder, but in the future a
high-penetration scenario of PV technology will require
facing a great variability created by a massive PV
distributed generation. An overall system needs to be
able to model at the same time the influence of large PV
plants and small distributed residential plants in terms of
grid voltages and frequencies disturbance preventing
large simultaneous phenomena. In fact, the possibility of
forecasting the future production of solar plants allows
enhanced planning and management of the energy
reserve, which leads, as an example, to the avoidance of
a forced shedding of power supply from renewable
energy plants, with a corresponding saving of energy
reserve correlated costs [9].
II.1.

1200

Meteorological Effects on PV Production

Variability

The output of PV plants is intrinsically variable

because of the sun path, but in addition clouds and
meteorological events are responsible for the
considerable changes in the output of PV plants and
these phenomena are often very fast.
In fact, a typical energy production curve is presented
in Figure 1, where are reported the power profiles of 3
days with different meteorological conditions: in
particular the first day is a sunny day, the second has
partial cloud coverage, the third shows the effects
(production drop) of a rapidly changing cloud coverage.

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved

10

20

30

40
Time (h)

50

60

70

Fig. 1. Measured solar radiation and produced power in a 1600 kW

plant under different meteorological conditions

Meteorological information about PV plants area can

be used to indicate approaching clouds and predict the
impact the atmospheric disturbance will cause on PV
output, as detailed in the following section. In fact, the
time it takes for a passing cloud to shade a PV system
depends on the PV plant size, cloud speed, and other
factors which can be predicted in some way [13].
Thanks to reliable weather forecasts, it is possible to
estimate in advance the expected quantities of the PV
energy being produced. Moreover the same weather
information is an important factor also in load
forecasting models and electricity demand behavior
estimation.
It has already been considered that, in a model of
massive PV energy distribution, clouds are able to cause
diverse changes and ramps in PV output in different
parts of the same plant or among different plants in the
same area. The degree of diversity among points or
plants can be characterized by the degree of correlation
of simultaneous changes in the output. The differences
between multiple PV sites need to be taken into account
in projections and actions that must be managed by
system operators. If the number of PV sources increase
substantially then the system complexity cannot be
neglected. However if the renewable energy sources are
sparsely deployed in a consistent way, the aggregation of
several diverse production sources leads to an aggregate
smoother profile than would be expected from scaling of
a single PV plant.
PV variability typically can manifest voltage or power
quality problems: to develop projections of PV
variability it is necessary to model on the time scale of
seconds to hours the output of large and dispersed PV
plants on distribution feeders [13].
II.2.

Soft Computing Tools for PV Production

Management

In order to foster a feasible development of

interconnection of PV to the electric power system,
International Review of Electrical Engineering, Vol. 7, N. 1

3455

Mikhail Simonov, Marco Mussetta, Francesco Grimaccia, Sonia Leva, Riccardo E. Zich

artificial intelligence tools can be effectively used to

enhance reliability of production forecast and thus
contribute to enhance reliability of a smart energy
system.
Recently, the authors have proposed one predictive
model based on soft computing techniques, serving the
independent PV plants [14]. In particular, considering
the complex nature of the forecasting problems, the use
of weather forecast processed by an Artificial Neural
Network (ANN) has been considered to solve them.
ANNs are effective computing methods, able to easily
process the dynamic behavior of time-varying variable:
they are particularly useful to model the complex and
nonlinear relationships among data, without any a-priori
assumption concerning the nature of these correlations.
In more practical terms neural networks are non-linear
statistical data modeling tools used to model complex
relationships between inputs and outputs or to find
patterns in data.

III. Artificial Intelligence Techniques

Implementation
In order to predict PV plants produced power, weather
forecasts are a key input for the ANN: for this reason, in
this paper a real-world case study with a fuzzy filtering

approach has been considered to validate the capability

of dealing with the complexity of inputs involved in the
available weather forecast data.
The generic computational model of an ANN
simulates the features and behavior of the human brain
neural networks. One of the most typical ANN
implementation extensively used in engineering
applications and in forecasting problems [15], [16] is the
Multi-Layered Perceptron (MLP). It consists of an
interconnected group of artificial neurons (as shown in
Figure 2) that suitably processes information according
to the strength of connections among them and to a given
non-linear activation function.
Training algorithms are used to train the network by
learning the nonlinear behavior between input and output
data. This learning process should find the weights
configuration associated to the minimum output error,
namely the weights configuration optimized to reach a
good and accurate desired output. Typically, the learning
process target is to minimize an error (or cost) function.
The cost function, in particular, is related to the error
between the network's output, f(x), and the target value y
over all the example pairs (i.e. mean-squared error). For
this reason it must contains prior knowledge about the
problem output.

Fuzzy logic pre-processing

noon forecast

morning forecast

in1

UV index

weights1

irradiance

cloud index

in2

weights2

in3
in4

Time

weights3

in5
in6
in7

out

in8

out

1
2

in9
in10

out

evening forecast

in11
in12
in13

bias 3

in14
in15

bias 2

in16
bias 1

Fig. 2. Detailed structure of the implemented feed-forward ANN with Fuzzy pre-processing

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved

International Review of Electrical Engineering, Vol. 7, N. 1

3456

Mikhail Simonov, Marco Mussetta, Francesco Grimaccia, Sonia Leva, Riccardo E. Zich

The learning task is the most critical part of an ANN

implementation, since it deals with an optimization
process covering a high number of independent
unknowns (the neural weights); as a consequence such a
search could be extremely difficult. Moreover, there
could be several weights configurations which are quite
similar each other and very close to the global optimum
case, even if these configurations are sub-optimal.
Although one of the typical approach used to optimize
and train a MLP is the Error Back-Propagation (EBP), a
gradient descent based strategy, recently the application
of Evolutionary Algorithms has been considered, e.g.
Genetic Algorithm (GA) or Particle Swarm Optimization
(PSO), in order to take advantage of their global
optimization capability.
In particular, the authors have already successfully
applied the Genetical Swarm Optimization (GSO)
algorithm [14], a hybrid search technique modeled on the
concepts of natural selection and evolution of GA and on
the social behavior derived from PSO.
More recently, the authors considered also the
integration of another soft-computing technique, namely
Fuzzy Logic, in order to increase the reliability of the PV
power forecast in estimating the solar irradiance [17]. In
fact, although during the bright days or heavy cloudiness
the PV forecast appears acceptable, still the described
soft computing solution gives an imprecise forecast for
some days.
Over the last decades fuzzy models have been
extensively used in engineering applications, especially
in control systems, easing the solution to many problems
by introducing an unconventional set of rules. System
analysis and control design for nonlinear systems based
on fuzzy system approaches have been very active in the
scientific community [18].
Here fuzzy models, due to their intrinsic capability of
approximating nonlinear data mappings, have been
exploited for the preprocessing of a sub-set of weather
data in order to filter the inputs for the ANN. In
particular, as shown in detail in Figure 2 (inset), some
meteorological indexes, namely UV index, cloud cover
and specific time/season, have been processed by a
Fuzzy procedure to increase the reliability of estimation
of the solar irradiance, since this parameter is the most
relevant to PV production but it is not directly provided
by the weather forecast service and must be modeled
separately [19].

IV.

Numerical Results

In order to validate the proposed approach, a real PV

plant production forecasting problem has been
considered, with different time horizons: short term (day
and several days), medium term (week and several
weeks), and long term (year and several years); forecasts
for different lead time can potentially be used for
different aims. We finally chose a short term time base
(one day in advance) for this specific application.

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved

After a preliminary trial campaign, presented in [14],

the network architecture which has provided better
results presents two hidden layers, as shown in Figure 2.
In particular, the number of neurons in the input layer
is 16, as reported in Table I which describes the
meteorological parameter provided from the weather
forecast service. Neurons i2,,i6 represent forecasted
values at 6am of the considered day (i1), neuron i7,,i11
at noon, while neurons i12,,i16 represent forecasted
meteorological parameters at 6pm.
TABLE I
CONSIDERED METEOROLOGICAL PARAMETERS
Neurons

Variable name

Range

i1
i2, i7, i12
i3, i8, i13
i4, i9, i14
i5, i10, i15
i6, i11, i16

Day
Irradiance
Air temperature
Wind speed
Air humidity
Air pressure

(1366)
(01350) W/m2
(030) C
(020) km/h
(0100) %
(10001040) mbar

The neurons in the two hidden layers are, respectively,

13 and 9, while the output layer can be composed of
several neurons, since different plants forecast can be
obtained simultaneously, provided that they are
sufficiently close to each other in order to share the same
input data, i.e. the weather forecast; in fact, as already
reported in [14], the authors performed preliminary
analyses for three different plants with different
productive capacity, and results showed that the
developed soft computing model behavior does not
change with respect to the plant power size.
Figure 3 shows the computed daily production
forecast compared with the actual measured data: this
simulations are based on a daily prediction, namely for
each day produced power data have been effectively
measured, while the corresponding weather forecast were
provided one day in advance.
To further increase the accuracy of the provided
forecast finding the best neural weights configuration,
here the hybrid evolutionary learning algorithm was
followed, in the last iterations, by a run of a suitably
developed EBP procedure: in fact, due to the high
number of neural weights to be optimized, this mixed
approach guarantees to speed-up the convergence of the
whole learning process.
The final behavior of the simulated data with respect
to the experimental measurements appears to be very
good because the average error is less than 2.3% over the
whole period of six months, significantly improving the
previously obtained value of around 5%.
An additional analysis of these results puts at a glance
that the critical points are mainly concentrated during the
days with cloud index in the interval {3-7}.
In such days, whose significance is low but not
negligible, as shown in Figure 4, the irradiance profile
changes suddenly due to clouds and their relative more
or less fast movement.

International Review of Electrical Engineering, Vol. 7, N. 1

3457

Mikhail Simonov, Marco Mussetta, Francesco Grimaccia, Sonia Leva, Riccardo E. Zich

Energy forecast (avg error = 2.2846%)

200

150

Actual production
Forecast
Current error (%)
Previous error (%)

100

50

100

120

140

160

180
Days

200

220

240

260

Fig. 3. Results of the forecasting over a six-months time frame and comparison with actual measured data.
Current %error level (in red) is compared with the previously obtained one, reported in [17]
Daily cloud coverage index
60
50

# days

40
30
20
10
0

4
5
Index value

Fig. 4. Distribution of daily cloud coverage index

in the considered 6 months time frame

To face this problem, giving in turn the improved

forecast precision, some additional heuristics capable to
reduce the negative effect on power quality are presented
below. This will contribute further in better PV
integration in smart grids.

V.

Future Perspectives for Increasing

PV Integration in Smart Grid

In case of rapid changes in solar radiation or

temperature at the forecasted day, produced power
greatly changes and forecast error would sensibly
increase. In traditional prediction methods the ANN uses
all similar days data to learn the trend of similarity.
However, learning all similar days' data does not help if
weather conditions change suddenly during the same
day. The negative and/or positive energy dynamics could
happen during the daytime because of clouds passing
over. The above changes satisfy the cause-effect
relationship because the mobility of the clouds is a
process evolving and propagating continuously over the
space.
Although the weather forecast can be locally provided
by different sources, including sensors and weather
stations installed, nowadays this information is not

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved

exchanged among the different energy production plants,

because of absence of awareness about each other.
However, when different power plants operate
simultaneously, the spatial scales of importance for
power quality may be in the range of
tens of kilometers [13].
Therefore, the presented neuro-fuzzy system could be
integrated with real time information coming from local
meteorological stations and, in particular, from
surrounding regions and cities, where the weather change
has already occurred [20]. In view of this, it is possible
to introduce the mutual knowledge between the PV
nodes resulting in the improved cooperation across a
smart grid model, likely giving in turn some improved
energy efficiency. The nodes PVi measure the energy
production levels and exchange the above knowledge
with the neighbours through the smart grid (e.g. using
digital energy messages). Let observe that the energy
dynamics happening at the node i having the nominal
power Pi, will be amplified at the node j having the
nominal power Pj where accordingly the ratio Pj(t+t) =
Pi(t)k, and k = Pi/Pj. This ratio indicates the potential to
perform the anticipatory control [20].
We can assume also a geographically distributed
topology composed by n photovoltaic generators PVi,
where i = 1, 2, , n. To catch energy dynamics, each
PVi node of the topology can be equipped by innovative
event-driven electronic metering device [21], and
connected to the electric grid supplying the renewable
energy quantities Ei(t). The above configuration
originates the energy flow being injected in the grid:
E (t ) =

Ei ( t )

(1)

i =1

For each source i, the value Ei(tk) becomes known in

the grid at t = tk (now), while the authors developed
ANN model, as described in [17], estimates future values

International Review of Electrical Engineering, Vol. 7, N. 1

3458

Mikhail Simonov, Marco Mussetta, Francesco Grimaccia, Sonia Leva, Riccardo E. Zich

Ei(tk+tk) in a probabilistic way, since it is based on

weather forecast. An energy drop Ei(tk) < Ei(tk-tk) might
happen at tk. The like oddness of the above event if
different, depending on weather parameters, e.g. the cited
cloud coverage index, with some examples already
known because of the knowledge based on the past series
already observed by the meteorological conditions. The
knowledge about the above phenomena becomes
available to the independent node i at tk. The lack of the
anticipatory knowledge about forthcoming production
drops is a drawback of the distributed generation by the
topology composed by independent nodes.
On a very short time horizon, during the days with
variable cloudiness, the weather forecast can be
corrected in real-time offering an anticipatory knowledge
about the unexpected energy production dynamics going
against the forecasted values, permitting to operate some
regulations before the energy increases or decreases
causing energy flow inversion.

References
[1]

[2]

[3]

[4]

[5]

[6]

[7]

VI.

Conclusion

Forecasts are an important tool for managing the

variability of PV plants and should be incorporated into
electric system planning and operations. As with other
renewable energy sources, solar forecasting can benefit
from further development of weather models and data
sets. Authors have proposed a Neuro-Fuzzy predictive
model to be used to reduce uncertainty during anomalous
days characterized by the variable cloudiness.
In this paper we discuss the potential optimization
degree locally achievable by extending the cooperation
in a future smart grid model, in which the collective
optimization is enabled by the anticipatory knowledge
about the energy production dynamics caused by the
cloud mobility during the variable cloudiness days. An
overall system needs to model at the same time the
influence of large PV plants and small distributed
residential plants to reach an improved grid stability.
Certain concerns regarding power quality and
regulation reserves will require in the future data with a
higher time resolutions. The advanced tools here
presented for analysis of data deriving from multiple PV
plants potentially will let solar energy from PV plants
reach more significant penetration levels.
The impact achievable in the real life depends also on
the capacity to operate the efficient decision making
within the correct timeframe. Future work will be target
to exploit the information available locally at the grid
level to be transformed into benefits in a collective
global optimization perspective.

[8]

[9]

[10]

[11]
[12]

[13]

[14]

[15]

[16]

[17]

Acknowledgements
This work was partially supported by Regione
Lombardia, ICT-Metadristretti Action, under grant
ID5109 - DGR N.6735.

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved

[18]
[19]

G. Brinkman, P. Denholm, E. Drury, R. Margolis, M. Mowers,

Toward a Solar-Powered Grid, IEEE Power and Energy
Magazine, vol. 9, n. 3, May-June 2011, pp. 24 32.
A. Moshari, A. Ebrahimi, A comprehensive model for zonal
short term load forecasting, International Review of Electrical
Engineering (IREE), vol.6, n. 1, January 2011, pp. 346354.
Y. Huang, J. Lu, C. Liu, X. Xu, W. Wang, X. Zhou, Comparative
study of power forecasting methods for PV stations,
International Conference on Power System Technology
(POWERCON), 24-28 Oct. 2010, pp. 16.
D. Fay, J. V. Ringwood, On the Influence of Weather Forecast
Errors in Short-Term Load Forecasting Models, IEEE
Transactions on Power Systems, Vol. 25, n. 3, August 2010, pp.
17511758.
H. S. Hippert, C. E. Pedreira, R. C. Souza, Neural networks for
short-term load forecasting: a review and evaluation, IEEE
Transactions on Power Systems, Vol. 16, No. 4, November 2001,
pp. 798805.
K. Methaprayoon, W. Lee, S. Rasmiddattaa, J. Lia and R. Ross,
Multistage Artificial Neural Network Short-Term Load
Forecasting Engine With Front-End Weather Forecast, IEEE
Industrial Applications, vol. 43, no. 6, pp. 1410- 1416, 2007.
Z. H. Ashour, M. A. Farahat, A new artificial neural network
approach with selected inputs for short term electric load
forecasting International Review of Electrical Engineering
(IREE), vol.3, n. 1, February 2008, pp. 32 36.
H. Zainab, M. Osman, M. L. Awad, T. K. Mahmoud, Neural
Network Based Approach for Short-Term Load Forecasting,
IEEE Power Systems Conference and Exposition, 15-18 March
2009, pp. 18.
E. Collino, C. Dainese, D. Ronzio, Application of solar radiation
forecast for the management of a mixed PV-biomass power plant:
a preliminary evaluation, EMS Annual Meeting Abstracts, Vol. 8,
EMS2011-530-1, 2011, 11th EMS / 10th ECAM, September 2011.
J. Frunt, W.L. Kling, J.M.A. Myrzik, F.A. Nobel, D.A.M. Klaar,
Effects of Further Integration of Distributed Generation on the
Electricity Market, Proceedings from UPEC 06: The 41st Intl.
Univ. Power Engineering Conf. 1 (2006), pp. 15.
Jiyuan Fan, S. Borlase, The evolution of distribution, IEEE
Power and Energy Magazine, 7(2) (2009), pp. 6368.
J.A. Pecas-Lopes, P.M. Rocha-Almeida, F.J. Soares, Using
vehicle-to-grid to maximize the integration of intermittent
renewable energy resources in islanded electric grids,
International Conference on Clean Electrical Power (2009), pp.
290295.
A. Mills, M. Ahlstrom, M. Brower, A. Ellis, R. George, T. Hoff,
B. Kroposki, C. Lenox, N. Miller, M. Milligan, J. Stein, Y. Wan,
Dark Shadows, IEEE Power and Energy Magazine, vol. 9, n. 3,
May-June 2011, pp. 33 41.
T. Senjyu, H. Takara, K. Uezato, Two-hour-ahead load
forecasting using neural network, Proceedings International
Conference on Power System Technology, 2 (2000), pp. 1119
1124.
W. Taylor, and R. Buizza, Neural network load forecasting with
weather ensemble predictions, IEEE Transactions on Power
Systems, 17(3) (2002), pp. 626632.
D. Caputo, F. Grimaccia, M. Mussetta, R.E. Zich, Photovoltaic
plants predictive model by means of ANN trained by a hybrid
evolutionary algorithm, Proceedings of the 2010 International
Joint Conference on Neural Networks (IJCNN), July 18-23 2010,
Barcelona, Spain.
F. Grimaccia, M. Mussetta, R.E. Zich, Neuro-Fuzzy Predictive
Model for PV Energy Production based on Weather Forecast,
Proceedings of the 2011 IEEE International Conference on Fuzzy
Systems, June 27-30, 2011, Taipei, Taiwan, pp. 2454 2457.
Ping-Teng Chang, Fuzzy seasonality forecasting, Fuzzy Sets
and Systems, Vol. 90, No. 1, 16 August 1997, Pages 1-10.
T. Krishnaiah, S. Srinivasa Rao, K. Madhumurth y, K.S. Reddy,
Neural Network Approach for Modelling Global Solar

International Review of Electrical Engineering, Vol. 7, N. 1

3459

Mikhail Simonov, Marco Mussetta, Francesco Grimaccia, Sonia Leva, Riccardo E. Zich

Radiation, Journal of Applied Sciences Research, vol. 3 n. 10,

2007, pp. 1105 1111.
[20] M. Simonov, Real time management of unpredictable renewable
solar energy in smart cities, Proceedings of the Solar Energy
Tech 2010, July 8, 2010, Milano, Italy, pp. 89 93.
[21] T. Singh, P.K. Vara, Smart Metering the Clouds, 18th IEEE Int.
Workshops on Enabling Technologies: Infrastructures for
Collaborative Enterprises (2009) , pp. 6671.

Authors information
1

Politecnico di Milano,
Dipartimento di Energia,
Via La Masa 34,
20156 Milano,
Italy.
2

ISMB Istituto Superiore Mario Boella,

Via P.C. Boggio 61,
10138 Torino,
Italy.
Mikhail Simonov. Ph.D. in Electrical
Engineering from Politecnico di Milano, Italy,
on Ubiquitous Near-Energy Services in
Extended Entities as an opportunity for Energy
Efficiency and Security related to smart grids.
M.S. in Applied Mathematics/ICT from Moscow
State University on artificial cognition related
topic. He has 20+ years of a multi-disciplinary
and cross-sectorial experience in ICT, Aerospace, Industrial
automation, and Finance. He attended a stage in management
consultancy at Bain, worked in Financial, Insurance companies, and
Software Industries. International project leader and manager for more
than 10 years. He performs applied R&D activities since 1999. He has
several peer-reviewed publications, assists EC as an independent
expert. M. Simonov is member of IEEE.
Marco Mussetta received the M.S. and Ph.D.
degrees in electrical engineering from
Politecnico di Milano, Italy, in 2003 and 2007,
respectively. Currently, he is an Assistant
Professor in the Energy Department of
Politecnico di Milano. His main research
interests are related to soft computing techniques
development and application in different fields,
such as electromagnetic devices, wireless sensor networks and
photovoltaic and other energy harvesting devices. Dr. Mussetta is
member of IEEE, PES, CIS and AEIT.

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved

Francesco Grimaccia received the M.S. and

Ph.D. degrees in electrical engineering from
Politecnico di Milano, Italy, in 2003 and 2007,
respectively. Currently, he is an Assistant
Professor in the Energy Department of
Politecnico di Milano. His main research
interests are related to soft computing techniques
development and application in different fields,
such as wireless sensor networks, photovoltaic and other energy
harvesting devices. Dr. Grimaccia is member of IEEE and AEIT.
Sonia Leva received the M.S. and Ph.D. degrees
in electrical engineering from the Politecnico di
Milano, Milano, Italy, in 1997 and 2001,
respectively. Currently, she is Associate
Professor of Electrical Engineering in the
Energy Department of the Politecnico di Milano.
Her research interests include electromagnetic
(EM) compatibility, power quality, the
foundation of the EM theory of the electric network, and the renewable
energy.
She is member of the Italian Standard Authority (CEI/CT82) and of the
IEEE Working Group Distributed Resources: Modeling & Analysis
and of Task Force on Modeling and Analysis of ElectronicallyCoupled Distributed Resources.
Riccardo E. Zich received the M.S. (summa
cum laude) and Ph.D. degrees in electronic
engineering from the Politecnico di Torino,
Torino, Italy, in 1989 and 1993 respectively.
He is now Full Professor of electrical
engineering in Politecnico di Milano. His main
research activities concern analytical and
numerical techniques in high- and lowfrequency
electromagnetics,
electromagnetic
compatibility,
evolutionary computation and non-conventional antenna and wireless
sensor networks design.
Dr. Zich is a member of the AP, EMC, MTT, and CAS societies, the
International Union of Radio Science (URSI) and the Institute of
Electronics, Information and Communication Engineers (IEICE).

International Review of Electrical Engineering, Vol. 7, N. 1

3460