DUAL FUNCTIONALITY ARTIFICIAL NEURAL NETWORK TECHNIQUE FOR WELLS TURBINE BASED WAVE ENERGY CONVERSION SYSTEM

Wave energy is of particular interest amongst new market-penetrating renewable energy resources. Although energy extraction from wave motion is still in its infancy, recent studies predict rapid development. Several studies have investigated various wave energy harvesting configurations. The oscillating water column (OWC) is the most common type. The OWC is based on a Wells turbine system coupled to a doubly fed induction generator (DFIG) grid connected via a back-to-back converter, which is inherently advantageous, enabling this arrangement to dominate the wave energy conversion field.
Like the majority of renewable energy sources, wave energy power must be carefully tracked in order to maximize efficiency, due to its non-linear relation with the differential pressure and the turbine speed. Several maximum power point techniques (MPPT) have been presented in the literature that varies in implementation complexity, tracking convergence, and fast tracking methods.
Despite several advantages offered by the OWC-based wave energy conversion system (WECS), the commonly installed Wells turbine suffers a critical phenomenon: stalling. During this operating condition, the turbine power is dramatically decreased causing a severe system disturbance, especially when grid integration is required. Various stalling avoidance techniques have been presented in the literature, including air flow control implementation and rotor speed control, in addition to variable-speed strategies. The air flow control-based techniques offer limited performance compared to rotor speed control.
This thesis develops a dual functionality enhanced-performance technique that avoids stalling phenomenon and maximizes the extracted power. A grid connected WECS Wells turbine coupled to DFIG, is mathematically modelled within. A robust decoupled active-reactive power control strategy is subsequently presented for grid connection purposes. A novel artificial neural network-based stalling avoidance and maximum power point tracking (MPPT) technique is proposed. The presented technique facilitates a simplified training procedure, adequate stalling avoidance, minimal grid power oscillations, fast convergence and wide turbine speed range operation. Rigorous simulation results using Matlab/Simulink software package, comparing the developed and classical techniques, are utilized to verify the presented technique effectiveness and superiority under operating conditions.