Soutenance mémoire
Nonlinear control
Recently nonlinear control techniques such as sliding mode control, backstepping and inputoutput feedback linearization have drawn interest in power electronics applications since they offer systematic, powerful and easy-to-implement methods (Marino and Tomei 1996). A nonlinear frequency and voltage control based on backstepping technique is developed for PV generator in (Okou, Akhrif et al. 2012). A multi-input multi-output MIMO nonlinear frequency and voltage control based on feedback linearization technique is developed for doubly-fed induction generators (DFIG) generator connected to synchronous generator (Sow, Akhrif et al. 2011). It is shown in these studies that the nonlinear control approach improves the general system performance in both transient and steady-state regimes since the exact nonlinearity of system is taken into account in control design. The lack of storage system in both techniques causes the renewable generator to operate below its maximum power point, MPP. The effectiveness of these nonlinear control approaches is highly dependent on the system parameters. On the other hand uncertainties in parameters such as load power, terminal voltage of SG, line inductance and SG, PV, DFIG model parameters affects the controller tracking.
Adaptive nonlinear control The design of adaptive control was introduced in 1950s. The first and most important applications of adaptive control were in mill industries in Sweden. Another important application of adaptive control has been the design of autopilots in flight control. The airplanes operate over a wide range of speeds and altitudes with nonlinear and time-varying dynamics. The different operating conditions of aircraft lead to the different unknown parameters in the system model. A sophisticated feedback control needs to be able to learn about parameter uncertainties. Two adaptive approaches were introduced in the literature; « direct and indirect » adaptive controls. In indirect adaptive control, the plant parameters are estimated online and used to calculate the control parameters. In direct adaptive control, the controller parameters are estimated directly without estimating the plant parameters (Krstic, Kanellakopoulos et al. 1995, Kaufman, Barkana et al. 1998, Åström and Wittenmark 2013). In a microgrid system with renewable energy integration, the system parameters change due to the load perturbation or the fluctuation in the intermittent power of renewable generator. Moreover there are some parameters in the system model which are unknown.
An adaptive control can improve the system performance by estimating the unknown parameters (Yazdani, Bakhshai et al. 2008). Some applications of adaptive control within a microgrid in recent years are listed as: the estimation of the grid frequency in a phase-locked loop (PLL) for active power filtering (Hogan, Gonzalez-Espin et al. 2014), the regulation of the common DC bus voltage with different renewable energy generators (Dragicevic, Guerrero et al. 2014), the adjustment of the weighted coefficients of active power-frequency droop (Li, Wang et al. 2015), the load sharing in a parallel-connected DC-DC converters in a DC microgrid (Augustine, Mishra et al. 2015), the protection and control (Laaksonen, Ishchenko et al. 2014) and the power balance during transition from grid-connected to islanding mode in a microgrid (Shi, Sharma et al. 2013). A nonlinear controller based on sliding mode control with adaptive voltage droop was proposed for a microgrid (Ferreira, Barbosa et al. 2013). The advantage of the adaptive nonlinear control is that it improves system behavior under both nonlinearities and uncertainties. To the best of our knowledge, no research has addressed the adaptive nonlinear control of a microgrid integrated with photovoltaic generators along with storage devices in the literature. Therefore it motivates us to investigate this subject in the next chapters of this thesis.
Robust adaptive nonlinear control
The adaptive laws and control discussed in previous section are designed with the assumption that the plant model is free of noise and disturbance. The designed controller is to be implemented on a practical system that is likely to differ from its mathematical and ideal models. The actual plant can be corrupted by noising measurement or any external disturbance. The discrepancies between the developed and real models may affect the system performance and robustness. The theory of robust adaptive nonlinear control was first presented by Kokotovic and Marino in 1991 (Kanellakopoulos, Kokotovic et al. 1991). A new robust adaptive nonlinear control based on backstepping scheme for frequency and voltage regulation was designed for DFIG wind turbine (Okou and Amoussou 2008). This strategy takes into account the uncertainty, disturbance and nonlinearity of the system in the control design. However the problems associated with the microgrid lacking the storage system (i.e. MPPT vs. frequency confliction) and inertia-less generators such as PV system (i.e. virtual inertia) are not addressed.
Overview of the recent maximum power point tracking approaches The low energy conversion efficiency of PV array hinders the widespread use of PV in power systems. In order to overcome this drawback, maximum power should be extracted from the PV system. This objective can be achieved by a MPPT which identifies the optimal operation of the PV systems. To date, several MPPT techniques have been reported which can be sorted into three categories; namely the conventional, soft computing and advanced model-based methods. Among conventional MPPT methods reported in the literature, the hill climbing (Elgendy, Zahawi et al. 2011, Ahmed, Li et al. 2012, Kumar 2012, Abuzed, Foster et al. 2014), perturb and observe (P&O) (Femia, Petrone et al. 2004, Liu and Lopes 2004, Femia, Petrone et al. 2005, Khaehintung, Wiangtong et al. 2006, Fangrui, Yong et al. 2008) and incremental conductance (IC) (Yuansheng, Suxiang et al. 2012, Guan-Chyun, Hung et al. 2013, Latif and Hussain 2014) are commonly used since they are quite simple to implement and they exhibit a good convergence speed. However, the oscillation around the MPP is the major drawback of theses algorithms (Banu, Beniuga et al. 2013).
The oscillatory behaviour around the MPP reduces considerably the system efficiency due to power losses. Moreover, when the atmospheric condition varies, these methods may be confused since the operating point can move away from the MPP instead of working around it. In order to minimize the oscillation, several attempts were made by reducing the perturbation step size. However, a smaller perturbation size affects the tracking speed adversely (Sera, Mathe et al. 2013, Shah and Joshi 2013). In order to improve the abovementioned drawback, soft computing (SC) techniques such as fuzzy logic control (FLC) (Ze, Hongzhi et al. 2010, Chin, Tan et al. 2011, Ze, Hongzhi et al. 2011, Arulmurugan and Suthanthira Vanitha 2013, Roy, Basher et al. 2014), Artificial Neural-Network (ANN) (Kaliamoorthy, Sekar et al. 2010, Phan Quoc,
Le Dinh et al. 2010, Pachauri and Chauhan 2014), genetic algorithm (Ramaprabha, Gothandaraman et al. 2011, Daraban, Petreus et al. 2013, Hadji, Gaubert et al. 2014, Mohamed, Berzoy et al. 2014), differential evolution (DE) (Taheri, Salam et al. 2010, Sheraz and Abido 2012, Taheri, Taheri et al. 2012), particle swarm optimization (PSO) (Kondo, Phimmasone et al. 2010, Phimmasone, Kondo et al. 2010) and firefly (FA) (Sundareswaran, Peddapati et al. 2014) algorithms have attracted much interest over the past years. One of the distinctive features of the soft-computing MPPT techniques comparing with other MPPT approaches is that they outperform in global searching during partial shading condition in PV system. Despite of their effectiveness, SC algorithms are more highly dependent on the complexity of computing programs (Paul 2013). In FLC MPPT, the membership function is generated through a time-consuming process. One of the major criticisms of ANN MPPTs is that they are considered as black boxes. Therefore no satisfactory explanation of their behaviour is offered. In stochastic techniques the decision variable, either the duty cycle of the power electronic converter or the reference voltage of the controller, is employed by the random vectors during the execution of algorithm. Therefore the global MPP convergence cannot be mathematically guaranteed.
|
Table des matières
INTRODUCTION
CHAPTER 1 LITERATURE REVIEW
1.1 Introduction
1.2 Overview of Microgrid Modeling
1.3 Overview of MG Control methods
1.4 Microgrid testbeds
CHAPTER 2 MATHEMATICAL MODELING
2.1 Introduction
2.2 Photovoltaic (Cell/module/array) model
2.3 DC-DC boost converter steady-state analysis:
2.4 Photovoltaic Converter Dynamics
2.5 Voltage Source Inverter (VSI) and Space Vector Modulation (SVM)
2.6 Inverter Voltage and Current Dynamics
2.7 Bidirectional Converter Voltage and Current Dynamics
2.8 Synchronous machine model
2.9 Frequency and voltage models
2.10 Nonlinear model of the entire system
2.11 Conclusion
CHAPTER 3 CLASSICAL CONTROL OF MICROGRID
3.1 Introduction
3.2 Microgrid System and Control
3.2.1 Inverter control
3.2.2 Battery converter control
3.2.3 Photovoltaic converter control
3.3 Results and Discussion
3.4 Conclusion
CHAPTER 4 NONLINEAR CONTROL OF MICROGRID
4.1 Introduction
4.2 Controller Design
4.2.1 Decoupling and Linearizing Control Law
4.2.2 Design of the Stabilizing Linear Control Laws
4.3 Simulation Results and Discussion
4.3.1 Frequency and Voltage Regulation in MG Islanding
4.3.2 Power Sharing Capabilities
4.4 Experimental Validation
4.5 Conclusion
CHAPTER 5 ROBUST ADAPTIVE NONLINEAR CONTROL OF MICROGRID
5.1 Introduction
5.2 System configuration and modeling
5.3 Proposed control scheme
5.3.1 Decoupling and feedback linearization control laws
5.3.2 Robust and adaptation laws with parameter estimation
5.3.3 Design of stabilizing linear control laws
5.4 Simulation Results
5.5 Conclusion
CONCLUSION
RECOMMENDATIONS
LIST OF REFERENCES
Télécharger le rapport complet
