ORGANIZER AND SPEAKER:

Dr. G K Venayagamoorthy, Senior Member, IEEE, FIET, FSAIEE, Member of the IEEE IES
Director: Real-Time Power and Intelligent Systems Laboratory
Department of Electrical and Computer Engineering, Missouri University of Science and Technology,
Rolla, MO 65409, USA
gkumar@ieee.org
http://brain2grid.org; http://mst.edu/~ganeshv; http://rtpis.org

GOAL:

The objective of this tutorial is to expose researchers from the academia and industry to advanced computational and learning methods for system identification/modeling, nonlinear control and optimization in a smart grid environment.

ABSTRACT:

This tutorial will focus on the following major topics, starting with introduction to the field of computational intelligence and its applications in system identification, control and optimization with emphasis on the smart grid environment.

• Conputational intelligence - Neural Networks; Neuro-Fuzzy; Evolutionary Neuro Systems; Swarm Intelligence; Artificial Immune Systems (AIS)
• Nonlinear System identification – Using neural networks and particle swarm optimization
• Intelligent Control - Adaptive and Optimal using neural networks, fuzzy control and AIS
• Dynamic Optimization using learning methods
• Smart Grid – What is a Smart Grid? Vehicle-to-grid technology; SmartParks; Wind integration; Wide area monitoring and control; Generator maintenance scheduling; voltage predictions; reactive power and voltage control; microgrids
System characterization and identification are fundamental problems in systems theory. The identification problem is to infer relationships between the past data and future ones of unknown time series and dynamical systems. In both cases the ultimate goal is to provide a time-dependent model approximating the behavior of the system generating the data. In classical approaches the search for the optimal approximation model is carried out within a parameterized identification family (usually linear such as Moving Average (MA), Auto-Regressive (AR), and/or their combination ARMA for time-series or with an eXogenous (X) variable e.g., ARX or ARMAX for dynamical systems) and it is chosen to optimize a given figure of merit (e.g. a mean-squared error function). Because of its simplicity, a linear model does not always adequately approximate a generic nonlinear system throughout its entire working environment. Therefore, to improve approximation accuracy, various solutions have been envisaged which generally encompass system linearization around a set of working points and expansions in series of functionals (e.g. Volterra’s series): the bottleneck with this is the computational load. Obviously, difficulties increase when the model ruling the system is unknown; in this case, the system needs to be treated as a black-box model.

For this purpose approximators like neural networks, fuzzy systems, because of their intrinsic nonlinearity and computational simplicity, are natural candidates to approximate a given model. In fact, the nonlinear parametric family obtainable with neural structures extends the linear one by nonlinear models; among them are the NAR and NARX subfamilies. Identification of dynamic systems with neural networks has been suggested for example by Narendra where, under the stationarity hypothesis for the system generating the data, it is shown how NARX neural networks are able to solve the problem. Feedforward and feedback neural networks are employed for systems identification. The feedforward multilayer neural networks are trained using the simple error backpropagation method whilst the recurrent network may be trained with its variant. There are several modifications to the use of neural networks, their training algorithms to achieve global minimum and, to retain long term and short term memory during offline and online trainings.

Nonlinear control has been proposed using intelligent techniques such as neural networks, fuzzy, reinforcement learning and many others using inverse models, direct/indirect adaptive, or cloning a linear controller. There are merits for each approach adopted. There is a wide-gap between applications of these methods in real time and in simulation. Issues such as stability, processor speeds, learning time, types of training algorithms etc. arise when it comes to real time implementations.

Adaptive Critic designs are neural networks capable of optimization over time under conditions of noise and uncertainty. The optimization technique is based on a combination of the concept of reinforcement learning and approximate dynamic programming. The Adaptive Critic method determines an optimal control law for a system by successively adapting two neural networks, an Action network (which dispenses the control signals) and a Critic network (which ‘learns’ the desired performance index for some function associated with the performance index).

Swarm intelligence is powerful algorithm based on the collective effort of a population of agents to accomplish a local or a global task. The communication among agents using the particle swarm optimization algorithm is fast and convergence of the algorithm on suitable selection of its parameters can be achieved. Optimal parameters can be selected using offline or online techniques.

The primary aim of this tutorial is to provide power, control and system engineers/researchers from industry/academia, new to the field of computational and learning methods with the fundamentals required to benefit from and contribute to the rapidly growing field of intelligent systems applications in the smart grid environment. In particular, a clear understanding of the different strategies for designing intelligent identifiers and controllers will be developed by means of examples for nonlinear systems.



BIOGRAPHY



Dr. Venayagamoorthy received his PhD degree in electrical engineering from the University of Natal, South Africa. Currently, he is an Associate Professor of Electrical and Computer Engineering, and the founder and Director of the Real-Time Power and Intelligent Systems (RTPIS) Laboratory at Missouri University of Science and Technology (Missouri S&T). He was a Visiting Researcher with ABB Corporate Research, Sweden, in 2007.

His research interests are in the development and applications of advanced computational algorithms for real-world applications, including power systems stability and control, smart grid applications and sensor networks. He has published over 360 articles in refereed journals and conference proceedings. He has been involved in approximately US$ 7 million of competitive research funding in the last seven years.

Dr. Venayagamoorthy is a recipient of several awards including a 2007 US Office of Naval Research Young Investigator Program Award, a 2004 NSF CAREER Award, the 2010 Innovation Award from St. Louis Academy of Science, the 2010 IEEE Region 5 Outstanding Member Award, the 2006 IEEE Power and Energy Society Walter Fee Outstanding Young Engineer Award, a 2007 Missouri S&T Teaching Commendation Award, a 2006 Missouri S&T School of Engineering Teaching Excellence Award, a 2008, 2007 and 2005 Missouri S&T Faculty Excellence Award and a 2009 Missouri S&T Faculty Research Award.

Dr. Venayagamoorthy is the Chair of the IEEE CIS SGTF on Smart Grid. He has been involved in the leadership and organization of many conferences including the Chair of the 2011 IEEE Symposium of Computational Intelligence Applications in Smart Grid (CIASG). Dr. Venayagamoorthy is a Fellow of the Institution of Engineering and Technology (IET), UK, and the South African Institute of Electrical Engineers. He is a Senior Member of the IEEE and the International Neural Network Society (INNS), and a Member of the American Society for Engineering Education. He is member of Board of Governors of INNS.