Assignment - Role of Computational Techniques On Proposed Research
Assignment - Role of Computational Techniques On Proposed Research
proposed research
Assignment I
Computational techniques
Ph.D Program (2013)
Submitted by:
Snehamoy Dhar,
Roll no. 13ELE014
Research fellow, Dept. of Electrical Engineering,
ITER, SOA University
I. Introduction:
In present scenario when in one hand abundance of fossil fuel storage is in global crisis and
on the other hand being a developing nation where per capita energy consumption is also
depicting of development, India is facing the challenge to facilitate electricity to its interior
population. The only solution to cope up with the situation is to accept and implement distributed
generation with micro grid in India. India is considering more than 80,000 villages with more
than 20 million people without electricity, according to 2012 energy survey. Many of those
villages are distant from utility grid. To implement new grid to connect the consumers to the
main distribution grid is not economic and thus beyond acceptability. The best choice is to
develop a microgrid utility to the population location with Distributed generation units (DERs).
These DERs are mainly targeted to get the power from non conventional energy sources (solar,
wind, bio fuel etc.). The problem with the non conventional energy sources is their inconsistent
nature. The load is also not constant with respect to time (different seasons in a year, population
increment in every year). Designing such a grid system with variable nature of sources as well as
loads, is quite complex and forecasting is one of the solutions to keep the operation simple.
Prediction of any data comes under the knowledge body known as Computational techniques.
All of the renewable sources are having non linear characteristics and to utilize the source with
maximum efficiency it is necessary to draw the load line through maximum power point (MPP).
To optimize the MPP various search algorithms are being implemented. These searching
techniques are also under the category of computational techniques.
II. Proposed research area:
Stable operation of a microgrid with several DER units, especially in an autonomous mode,
requires a power management strategy (PMS) and an energy management strategy (EMS). Fast
and efficient response of the PMS/EMS is more demanding for a microgrid compared with a
conventional power system. The reasons are:
Presence of multiple, small DER units with significantly different power capacities and
characteristics.
Potentially no dominant source of energy generation during an autonomous mode; i.e., lack
of infinite bus.
Fast response of electronically coupled DER units that can adversely affect voltage/angle
stability when appropriate provisions are not in place.
The PMS/EMS assigns real and reactive power references for the DER units to:
Appropriately share real/reactive power among the DER units.
Appropriately respond to the microgrid disturbances and transients.
Determine the power set points of the DER units to balance the microgrid power and restore
the frequency.
Enable resynchronization of the microgrid with the main grid, if required.
Figure 1 shows information/data flow and functions of a PMS/EMS for a microgrid. The real-
time management block receives the present and the forecasted values of load, generation, and
market information to impose appropriate controls on power flow, output generation, and
consumption level of the utility grid, dispatchable sources, and controllable loads, respectively.
Figure 1. Information/data flow and functions of a PMS/EMS for a microgrid
For power quality/ energy management of microgrid based on non-conventional DERs required
forecasted data for unit price determination and grid synchronous operation. The forecasting is
mainly for wind energy, solar radiation based on data collected throughout the year or past 5
years. Then these forecasted data compared with load forecasting data. The management strategy
is used to take decision for efficient and continuous operation with/ without grid supply (Grid
interactive/ Autonomous).
III. Involvement of computational techniques:
3.1. Different forecasting methods:
A. Artificial Neural Network (ANN):
An ANN is an information processing system that has certain performance characteristics in
common with biological neural networks. ANNs have been developed as generalization of
mathematical models of human cognition or neural biology based on some assumptions.
It has the characteristics like: (1) Its pattern of connection between the neurons (called its
architecture); (2) Its method of determination of weight on the connections (called as training or
learning algorithm); (3) its activation function to output the weighted inputs. For example:
Figure 2. RBF network architecture with single output unit
Figure 2 is a Radial basis function (RBF) network which is one type of ANN. RBF) networks
have also proven to be universal approximator, similar to the neural network and fuzzy interface
system. Because of the simple and fixed three-layer architecture as shown in figure 1, it is
effortless to design and train RBF networks than neural networks. From the point of
generalization, RBF networks can respond well for patterns which are not used for training. RBF
networks have strong tolerance to input noise, which enhances the stability of the designed
systems. Therefore, it is reasonable to consider RBF network as a competitive method of
nonlinear controller design. Architecture of any neural network is divided into two important
parts: (a) network construction; (2) parameter adjustment by training algorithm. RBF networks
are in the multilayer perceptron (MLP) category because of their layer-by-layer topology.
a) Network construction:
The RBF network designed with I number of input layer, H number of hidden/ kernel layer
and single output layer as shown in figure 1. The network is distribution is described in three
subdivisions:
1) Input layer construction:
A set of input data,
,1 ,2 ,3
, ,
{ , , ,..., ,..., }
p p p p
p i p I
x x x x x x = is applicable in the input layer. Each
input value of the input data set
, p i
x is further scaled by the corresponding input weight
, i h
u
presents in between i th input layer and hth hidden/kernel layer. For simplest approach input
weight may be considered as 1. Hence, the input weight matrix could be a set of values of I x H
dimention; where, I is input data set column dimension and H is number kernel units. The final
weighted input to the network is represented by
, , ,1 , ,2 , ,3 , , , ,
{ , , ,..., ,..., ,}
p h p h p h p h p h i p h I
y y y y y y = ;
where
, , , p h p i i h
y x u = (1)
2) Hidden/ Kernel layer construction:
In each hidden units an activation function is defined to calculate the output of the unit. In
proposed RBF network one Gaussian function is fitted as:
2
,
( ) exp
p h h
h p
h
y c
x
| |
| =
|
\ .
(2)
where
h
c and
h
center and width respectively of the Gaussian activation function ( )
h p
x for
hth RBF unit. - represents the Euclidean norm calculation of two vectors.
3) Output layer construction:
The output layer in proposed system consists of one unit only. For more complicated and
parallel computation more than one unit is appreciable. Here the output unit is functioned to
calculate sum total of all weighted, activation function generated inputs. The output weight
vector of
h
w scales the RBF unit outputs and calculates output
p
o as shown:
0
1
( ).
H
h p p h
h
o x w w
=
= +
(3)
0
w is bias weight.
It is important to set correct initial states for RBF networks before training.
b) Parameter adjustment by training algorithm:
The error back propagation (EBP) algorithm [Rumelhart et al. 1986] faded from the field of
artificial neural networks (ANN). But EBP was one of the most compelling developments in
neural network training. However, it is also known as an improper algorithm for present scenario
because of its slow convergence. Many research have been taken place to overcome the problem
of EBP algorithm and some highlighted are momentum and RPROP algorithm. As because of
being first order algorithm, the abovementioned didnt bring any drastic change until second
order algorithm, such as Newton algorithm and Levenberg Marquardt (LM) algorithm. They use
Hessian matrix to perform better estimations on both step sizes and directions, so that they can
converge much faster than first order algorithms. Based on gradient decent concept, many
algorithms have developed to perform accurate training on RBF networks, by not only adjusting
output weights but other important parameters, such as centers and widths of the activation
function for RBF units during the training process. First-order gradient method has limitations on
search ability and long convergence time requirement. By proposing a newly improved second
order (ISO) gradient algorithm the mentioned disadvantages are tried to be overcome.
B. Fuzzy Logic system:
In classical, or crisp, sets the transition for an element in the universe between membership
and non-membership in a given set is abrupt and well-defined (said to be crisp). For an
element in a universe that contains fuzzy sets, this transition can be gradual. This transition
among various degrees of membership can be thought of as conforming to the fact that the
boundaries of the fuzzy sets are vague and ambiguous. Hence, membership of an element from
the universe in this set is measured by a function that attempts to describe vagueness and
ambiguity.
A notation convention for fuzzy sets when the universe of discourse, X, is discrete and finite,
is as follows for a fuzzy set A:
(4)
Elements of a fuzzy set are mapped to a universe of membership values using a function-
theoretic form.
Figure 3. Membership function of Fuzzy set A~.
C. Other Computational techniques:
For particular DER unit for example photovoltaic (PV) or wind turbine based systems MPP
tracking is required as mention early. The nonlinear nature of output from solar irradiation and
wind flow is required to track maximum power point and compel the system load to operate on
that point. Some of the efficient tracking algorithms are hill climbing algorithm (thumb rule of
maximum fixed point), perturb and observe method (step wise perturb the data from nonlinear
curve and observe with previous data), Incremental conductance (maximize the nonlinear
function by finding 0
dy
dt
= , where y is the function of nonlinear curve), Fibonacci search
algorithm, Fuzzy logic based / ANN based algorithms.
IV. Conclusion:
Computational techniques are providing the way to model, design and simulate any hypothetical
solution in mathematical equation based formulation for any branch of knowledge. Before
practical implementation of any information testing of model in computer based simulators
become obvious and all the credits are for computational technique methods. In present energy
scenario in India it is highly require to implement distributed energy system with microgrid
based projects for consumers in rural areas. To utilize the maximum efficiency from any DER
unit, MPP tracking is highly solicited. Grid synchronous, continuous, efficient operation of
microgrid required forecasted data of sources as well as loads. All these process requires highly
efficient computational techniques to model the system.
References:
[1] Dasgupta, S. ; Sahoo, S.K. ; Panda, S.K., A new control strategy for single phase series connected PV module
inverter for grid voltage compensation, International Conference on Power Electronics and Drive Systems,
2009. PEDS 2009, 10.1109/PEDS.2009.5385813, pp: 1317-1322.
[2] Fei Kong, Rodriguez Cuauhtemoc, Amaratunga Gehan, Panda Sanjib Kumar, Series Connected Photovoltaic
Power Inverter, IEEE international conference on sustainable energy technology, 2008, ICSET 2008, pp595-
600, Nov 24 -27.
[3] Ming Tsung Tsai, Design of a Compact Series-connected AC Voltage Regulator with an Improved Control
Algorithm, IEEE transactions on industrial electronics, Vol 51. No. 4, pp 933-936, August 2004.
[4] Dasgupta, S. ; Sahoo, S.K. ; Panda, S.K., A novel current control scheme using Lyapunov function to control the
active and reactive power flow in a single phase hybrid PV inverter system connected to the grid, International
Power Electronics Conference (IPEC), 2010, 10.1109/IPEC.2010.5542125, Publication Year: 2010 , pp: 1701-
1708.
[5] Richard Zhang, Mark Cardinal, Paul Szczesny, Mark Dame, A grid simulator with control of single-phase
power converters in D-Q rotating frame, 33rd IEEE Annual Power Electronics Specialists Conference, 2002,
pesc 02, Vol. 3, 23-27 June 2002, pp.1431 - 1436.
[6] V. Khadkikar, A. Chandra, B. N. singh, Generalised singlephase p-q theory for active power filtering:
simulation and DSP - based experimental investigation, IET transaction on Power Electronics, Vol. 2, No. 1,
pp. 67-78, 2009.
[7] Jiabing Hu ; Bin Hu., Direct active and reactive power regulation of grid connected voltage source converters
using sliding mode control approach, Symposium on Industrial Electronics (ISIE), 2010 IEEE International,
10.1109/ISIE.2010.5637630 Publication Year: 2010 , Page(s): 3877 - 3882.
[8] Takahashi and T. Noguchi, A New Quick-Response and High- Efficiency Control Strategy of an Induction
motor, IEEE Trans. Industry Applications, Vol.22, No.5, 1986, pp. 820-827.
[9] M. Depenbrock, Direct Self-Control (DSC) of Inverter-Fed Induction Machine, IEEE Trans. Power Electronics,
Vol.3, No.4, 1988, pp. 420-429.
[10] T. Noguchi, H. Tomiki, S. Kondo and I. Takahashi, Direct Power Control of PWM Converter without Power-
Source Voltage Sensors, IEEE Trans. Industry Applications, Vol.34, No.3, 1998, pp. 473-479.
[11] D. Zhi, L. Xu, and B. W. Williams, Improved Direct Power Control of Grid-Connected DC/AC Converters,
IEEE Trans. Power Electronics, Vol. 24, No. 5, May 2009, pp. 1280-1292.
[12] Hooshmand, H.A. Malki, J. Mohammadpour, Power flow management of microgrid networks using model
predictive control, Computers & Mathematics with Applications, Volume 64, Issue 5, September 2012, Pages
869-876.
[13] Orecchini F, Santiangeli A., Beyond smart grids The need of intelligent energy networks for a higher global
efficiency through energy vectors integration., Int J Hydrog Energy 2011, 36: 812633.
[14] Wu DW, Wang RZ., Combined cooling, heating and power: a review, Prog Energy Combust Sci 2006;32:459
95.
[15] Lin L, Wang YD, Al-Shemmeri T, Ruxton T, Turner S, Zeng SC, et al., An experimental investigation of a
household size trigeneration, Appl Therm Eng 2007;27:57685.
[16] Angrisani G, Rosato A, Roselli C, Sasso M, Sibilio S., Experimental results of a micro-trigeneration
installation, Appl Therm Eng 2012;38:7890.
[17] Manisa Pipattanasomporn, Hassan Feroze, Saifur Rahman, Securing critical loads in a PV-based microgrid
with a multi-agent system, Renewable Energy, Volume 39, Issue 1, March 2012, Pages 166-174.
[18] Nwana HS. Software agents: an overview. The Knowledge Engineering Review, 1996;11(3):205e44.
[19] Bradshaw JM. Software agents. Cambridge, MA: MIT Press; 1997.
[20] McArthur SDJ, Davidson EM, Catterson VM, Dimeas AL, Hatziargyriou ND, Ponci F, et al. Multi-agent
systems for power engineering applicationsdPart I: concepts, approaches, and technical challenges. IEEE
Transaction on Power Systems 2007;22(4):1743e52.
[21] McArthur SDJ, Davidson EM, Catterson VM, Dimeas AL, Hatziargyriou ND, Ponci F, et al. Multi-agent
systems for power engineering applicationsdPart II: technologies, standards, and tools for building multi-agent
systems. IEEE Transaction on Power Systems 2007;22(4):1753e9.
[22] Lagorse J, Paire D, Miraoui A. A multi-agent system for energy management of distributed power sources.
Renewable Energy 2010;35:174e82.