IEEE Power & Energy Magazine - May/June 2015 - 89

Network Monitoring and Automation
and Analysis Functions
Grid resiliency depends on many factors, but a fundamental
element is being able to forecast what is going to happen in
advance. In this sense, having good information through a
monitoring system is basic. The second component would
be a reliable automation system, supported by analysis functions such as state estimators, optimal power flow tools, and
congestion forecasting.
The Spanish demonstration project, PRICE, includes several experiences related to system monitoring. Usually the
distribution grid is operated by monitoring and controlling
HV and MV levels, but LV ones are not monitored. Due to
the increasing DRES penetration, this is changing toward a
higher and more detailed visibility.
There are two questions to be answered. The first question could be whether it is really necessary to monitor the
LV grid or not, and the second question, not only applicable
to LV, would be which is the optimum level of monitoring.
To answer the first question, LV monitoring equipment
has been deployed in the area of Corredor del Henares
(Figure 5), taking advantage of the smart metering installation campaign. The main LV measurements in SSs are
collected and sent to a central system sharing the telecommunication infrastructure of metering. The added value of monitoring this information is being studied in several initiatives.
The information about the electricity production of distributed generators is not usually available for DSOs, but it
would be useful for distribution purposes. The project is collecting this information from generators in this area. An algorithm called an energy balancing system has been developed
to take into account the data from generators and SSs with
the grid topology for local balancing purposes at the SS level.
To answer the second question, this area has been partitioned in different levels of data collection density to evaluate the optimal level of information, considering not only the
number and duration of disruptions in each area but also the
incremental cost. An area where the original level of automation has been 3.8% of the contracted power changed to
10% of automation and 36% of supervision with a reduction
of System Average Interruption Duration Index (SAIDI) of
63%. A different one changing from 2.2% of automation level
to 5.5% with 20% of monitoring showed a reduction of 40%
in SAIDI. Those are preliminary results that depend on certain uncontrolled variables, but the first results are promising.
In the Greek demonstration, stochastic RES (PV) forecasting and probabilistic load flow are used to implement
advanced functions such as congestion management, RES
hosting capacity estimation and management, power quality monitoring, RES condition monitoring, and reduction of
network losses. This set of distribution management tools is
based on the automatic meter reading infrastructure.
The function of congestion management uses PV and
load forecasting and calculates the probabilistic load flow
(PLF), giving information about the sensitive nodes and
may/june 2015

the specific PV installations and loads, which can be controlled or reconfigured to avoid overloads. Using stochastic
short-term (12-24 hours ahead) PVs and load forecasting
and applying PLF techniques, the operator is warned of
expected congested lines with a certain probability. Using
sensitivity techniques, he can be alerted about critical PV
injections or sensitive loads that can be potentially controlled, thereby avoiding a complete disconnection of large
parts of the network.
As input, the system uses time series of power, numerical weather predictions, and static PV data (such as installed
capacity), and advanced forecasting models based on radial
basis function neural networks are used to create estimates
of future RES production including confidence intervals relevant to the accuracy of estimation. The numerical weather
predictions (Figure 6) comes from the SKIRON model, a
high-resolution meteorological model that provides a fiveday-ahead forecasting horizon, once per day.
The model uses a novel neural network for the estimation
of the solar power, based on the ARTMAP networks (predictive adaptive resonance theory) and the multiscale radial basis
neural network. It also uses a genetic algorithm to estimate
its parameters, not needing more than 10 min for an update.
This means that the proposed model can be implemented for
the short term such as very-short-term forecasting. The model
manipulates weather predictions without any historical data
or additional time for offline training. It runs once per day and
provides solar power forecasts for the next four days.

For Further Reading
IGREENGrid. (2015). [Online]. Available: www.igreengridfp7.eu
More Micro-grids EU Project. (2009, Dec.). DG3&DG4.
Report on the technical, social, economic, and environmental benefits provided by microgrids on power system
operation. [Online]. Available: http://www.microgrids.eu/
documents/668.pdf
metaPV EU Project. (2011, Dec.). D3.3-Active and autonomous operation of networks with high PV penetration.
[Online]. Available: http://www.metapv.eu/sites/default/
files/PR_PR104383_D3.3%20Active_and_autonomous_
operation_F.pdf

Biographies
Jesús Varela is with Iberdrola Distribución, Madrid, Spain.
Nikos Hatziargyriou is with the National Technical University of Athens, Greece.
Lisandro J. Puglisi is with Iberdrola Ingeniería y Construcción, Madrid, Spain.
Gareth Bissel is with Enel Distribuzione, Italy.
Andreas Abart is with EAG, Austria.
Marco Rossi is with RSE, Milan, Italy.
Robert Priewasser is with Salzburg AG, Austria.
p&e
ieee power & energy magazine

http://www.igreengrid http://www.fp7.eu http://www.microgrids.eu/ http://www.metapv.eu/sites/default/

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - May/June 2015