Adaptive filtering and control algorithms for grid integration of photovoltaic systems

Among several renewable energy sources, power generation from Photovoltaic (PV) systems has achieved a tremendous growth in the last few years due to diminishing costs of PV modules in addition to their increased efficiency. Usually, most of the largely powered PV systems are configured to be operated in grid-connected mode to achieve effective usage of solar PV power. PV System connected to Utility Grid (PVS-UG) can either be operated in single-stage or double-stage mode depending upon the power electronic conversion stages involved. In a single-stage PVS-UG, power loss is less compared to a double-stage mode due to absence of dc-dc converter stage. Therefore a single-stage mode of operation is preferred and is considered in the current research work. However, a number of control challenges are encountered while synchronizing a PVS-UG. The rise in use of nonlinear loads has resulted in harmonics injection and other power quality (PQ) issues in the distribution system. Further, the intermittent nature of solar energy is also a major challenge for a PVS-UG. In face of handling dynamic conditions related to load and environment along with the PQ issues, this dissertation focuses on developing adaptive filter based multi-functional control schemes for efficient functioning of a three-phase single-stage PVS-UG. Firstly, an adaptive filter based Modified Leaky Least Mean Square (MLLMS) control scheme has been developed for a PVS-UG. In the proposed MLLMS-based control scheme, an Incremental Conductance (InC) Maximum Power Point Tracking (MPPT) algorithm is used for maximum power extraction, MLLMS algorithm for extraction of fundamental active and reactive components of the load current, delivering PV power to the grid, balancing the grid current and compensating harmonics of the connected loads at the Point of Common Coupling (PCC). By introducing a leakage factor and selection of sum of exponential of the adaptation error in the cost function, the MLLMS algorithm overcomes the problem of drifting, low convergence and oscillations in weights encountered by some popular adaptive algorithms, e.g., Least Mean Square (LMS) and Least Mean Fourth (LMF) algorithms. The proposed MLLMS based control scheme is simulated in MATLAB/Simulink under load unbalance and irradiance change conditions. Subsequently, the said control scheme is realized on a prototype PVS-UG developed in the laboratory. From both the simulation and experimental results, it is observed that the proposed MLLMS based control algorithm outperforms LMS and LMF based control schemes in terms of reduced values in mean square error, oscillation in weights and Total Harmonic Distortions (THDs) of the grid currents. It is necessary to design robust controllers to achieve the grid connected PV system to perform even better during sudden changes in load and environmental conditions. A novel Variable Step Size Robust Least Mean Least Square (VSS-RLMLS) algorithm based control scheme is then designed for a PVS-UG. By implementing a generalized logarithmic cost function which combines higher and lower order measures of error in the weight update function, the proposed VSS-RLMLS algorithm achieves superior performance. In this control scheme, the VSS-RLMLS algorithm estimates the fundamental components of the nonlinear load currents accurately even during dynamic conditions thus allowing efficient functioning of the Voltage Source Converter (VSC) thereby delivering high quality power to the grid and mitigating the harmonics of the loads. From the simulations and experiments performed the proposed VSS-RLMLS based control scheme is seen to outperform control schemes employing LMF, variable leaky LMS and MLLMS in terms of better convergence, less steady state error, robustness during dynamic conditions and less oscillations in weights. The improvement in PQ is confirmed from lower grid current THD which is well within the IEEE-519 standards. Another robust adaptive filter based on the logarithmic cost function is the Least Logarithmic Absolute Difference (LLAD) algorithm. By embedding the conventional cost function of LLAD algorithm into the sigmoidal framework, the Sigmoid LLAD (SLLAD) algorithm is developed which further improves the performance of logarithmic cost function based robust adaptive filters. This new cost function based on the sigmoidal framework exploits the saturation characteristics of the nonlinearity of sigmoid function to achieve the improvement in performance. In the proposed control scheme, the SLLAD control algorithm estimates the fundamental components of nonlinear load currents accurately during dynamic conditions which results in generation of accurate reference grid current. From the simulation and experiments performed the proposed SLLAD based controller is seen to outperform controllers that employ LMF, MLLMS and LLLAD in terms of improved robustness, faster convergence, less steady state deviations and less oscillations in weights. The improvement in PQ is confirmed from lower grid current THD which is well within the IEEE-519 standards. From the comparative assessment it is found that all the proposed control schemes perform well during dynamic changes in the load and environmental conditions maintaining the grid currents sinusoidal, balanced and at unity power factor with THD of grid currents within the IEEE-519 standards. However, it is found that the SLLAD based control scheme performs the best among the above three control schemes during sudden changes in load and environmental conditions with less oscillations in weights and less steady state deviations. As a result, the THD of the grid currents using SLLAD based control scheme is also found to be the lowest among all control schemes. Thus, it is concluded that the SLLAD based control scheme exhibits superior performance among all the aforesaid proposed control schemes