Student  Name  :   Hessa Jassim Al-Junaid            

University of Bahrain,  Department of Electrical and Electronics Engineering,    P.O. Box 32038, Isa Town,  Kingdom of Bahrain.

Abstract

Hinf  robust fuzzy control systems have received a substantial amount of attention in the control research.  This is due to the potential benefits of mixing a knowledge-base system representing the “fuzzy systems” with the well-established theorems in the area of “Hinf robust control”.   This has resulted in a knowledge-based robust control system that is governed by rules in terms of describing the control aspects of a system operation.   In this sense, this thesis has presented two main issues related to Hinf robust fuzzy control of nonlinear dynamical systems.  The first has been fuzzy modeling of nonlinear dynamical systems, whereas the second was directed towards Hinf fuzzy gain-scheduling control systems.   Regarding fuzzy modeling, that was achieved by employing Takagi-Sugeno (T-S) fuzzy modeling technique.  The employed (T-S) fuzzy modeling technique was able to cluster the entire nonlinear global model into linear sub-models that are governed by fuzzy rules.  (T-S) fuzzy modeling has a number of potentials with respect to other fuzzy modeling techniques; that is a small number of rules are obtained even for complicated multivariable systems.   With respect to the Hinf fuzzy gain-scheduling, the thesis has first presented an approach for designing Hinf fuzzy controller for disturbance rejection via defining a Lyapunov potential function of the system fuzzy model, hence reducing the problem to a standard Linear Matrix Inequalities (LMI) formulation.   Hinf fuzzy gain-scheduling has been achieved via treating the (T-S) fuzzy sub-models as a Linear Parameter Varying (LPV) system, hence synthesizing a scheduling controller for variation in parameters, while preserving the robustness character of the controller in phase of parametric variation.   Hinf robust fuzzy control synthesis and Hinf fuzzy gain-scheduling have been both verified for two nonlinear systems; a SISO mass-spring-damper system and a MIMO antenna system.

Student  Name  :   Amin  Mohammed Sultan

University of Bahrain,  Department of Electrical and Electronics Engineering,    P.O. Box 32038, Isa Town,  Kingdom of Bahrain.

Abstract  

This research concentrates on the issue of employing computational intelligence dynamic model structures for identifying nonlinear plant used in constrained nonlinear model predictive control. In this respect, two approaches have been followed to model a single input single output (SISO) system and a multi-input multi-output (MIMO) system.  Both systems exhibit nonlinear behaviors and input constraints.   For the first approach, we concentrated on employing Artificial Neural Networks with nonlinear activation function and nonlinear training to model nonlinear plants used in model predictive control system with constraints. The employed algorithm has shown reasonable result of convergence; in addition to a good neural structure for modeling the nonlinear behavior of plants under control. For these neural models, the model predictive control problem has been formulated as a standard nonlinear quadratic programming problem with nonlinear constraints. The built neuro model predictive control has shown a great deal of controller synthesis even under nonlinear plant operation with input constraints.   In the second approach, we concentrated on employing Neurofuzzy models that operate over linearized nonlinear plant regions. In this respect, neurofuzzy models have been synthesized and employed to model linearized dynamics of plants with linear model predictive control strategy subject to input constraints. Neurofuzzy model predictive control has shown good ability to operate under linearized plant models. Hence linear model predictive control with input and output constraints controller structure can be synthesized.    Finally, the research frame work has concentrated in comparing the two employed approaches, for the nonlinear model predictive control. 

Student  Name  :   Ashraf Yousif Abo-Shanab   (Co-supervision)

  4- Solar Energy – Machine Torque Optimization Via Fuzzy Logic

Student  Name  :   Mohammed  Yousif  Al-Hammad   (Co-supervision)

University of Bahrain,  Department of Electrical and Electronics Engineering,    P.O. Box 32038, Isa Town,  Kingdom of Bahrain.

Abstract  

The objectives of this study are :

Maximizing the power transferred from photovoltaic generator (PVG) to the water pumping system, which utilizes a 3-phase induction motor.   Including the effect of solar cell temperature on the PVG and hall drive system.  Introducing of analytical controller to achieve best matching.  Establishment of a relation between maximum powers delivered from PVG, inverter frequency and induction motor for batter matching.  Improve the matching by selecting best inverter frequency values  to train the neurofuzzy mapping relation and control the inverter frequency.  Introducing two sensors technique to measure the maximum power of the PVG array.  Introducing neurofuzzy estimation to estimate the maximum power of the PVG array. 

5-  Power Demand Predication Via Evolving Neural Network Structures

Student  Name  :   Ali  Abdulla Al-Jamea    (Co-supervision)

University of Bahrain,  Department of Electrical and Electronics Engineering,    P.O. Box 32038, Isa Town,  Kingdom of Bahrain.

Abstract  

This research has been carried out as a master project,  and has lasted for four months. In its contexts,  the project has been divided into two main areas of interest related to the State of Bahrain power system planning and generation.  The project first area of research,  has been the Reliability Assessment and Analysis for the load demand for the State of Bahrain.  This has been analyzed through: Expected Energy Not Supplied (EENS), Frequencies of Load Occurrences,  Probabilities of Load to Occur  calculations.   The second research area of interest was concentrated on Intelligent Based  Electrical Power Demand Predication (IEDP) for the state of Bahrain.  In this context, IEDP consisted of Neural Network Predication as compared to Genetically Evolved Neural Predication mechanisms.   Hence specific reliability analysis have been conducted on  both presented  IEDP approaches. In the research,  it has been strongly concluded, that Genetically Evolved Neural Networks are robust and accurate mechanism of predication   compared to Neural Networks Prediction. This is due to the genetically evolved weights  and biases as related to the back-propagation learned neural network predictor.   In addition to this,    the learning mechanism is based on genetic programming,  which takes care of non-smooth power demand nature.

6- Genetically Controlled  Variable Structure  Induction Machines Control,  ( PhD   thesis  Co-supervision).