Adaptive Reconfigurable Photovoltaic Arrays Based on Spatially Dispersed Irradiance Profiles

This thesis intends to develop a modeling method for an increased-efficiency adaptive reconfigurable photovoltaic (PV) system. Power generation in PV systems is directly proportional to the incident solar irradiance values on PV module surfaces. High dependency of PV power generation on the incoming solar irradiance values has led the research to develop various models for irradiance estimation purposes. A good model from PV applications point of view should be able to take into account the clouds and their light interaction characteristics in order to generate reliable site-specific irradiance profiles or time-series. This thesis proposes a model for generation of Spatially Dispersed Irradiance Profiles (SDIPs) which utilizes real cloud patterns derived from sky images taken at the application sites and takes into account different cloud types and distributions in the sky together with their sunlight interaction characteristics in order to generate instantaneous irradiance profiles as well as daily irradiance time-series. Utilization of a comprehensive set of cloud types and their sunlight interaction characteristics by the model allows for a precise analysis of the effects of clouds on the incoming irradiance values. Each PV module on the Earth`s surface receives the beam irradiance through circles/ellipses formed when the existing cloud layer in the sky cuts the cones through which the PV module sees the Sun`s disk. The thickness of the circles/ellipses vary according to the variations of the position of the Sun in the sky during a day. The model assumes that the beam irradiance received by each PV module is only affected by the amount of cloud coverage enclosed within the mentioned circles/ellipses. The most appropriate circles/ellipses are determined for each PV module at each time instant and virtually located on the sky images. The beam irradiance component for each PV module is then obtained as a result of taking into consideration the attenuating effects of the cloud coverage enclosed within the mentioned circles/ellipses. Since the thickness of circles/ellipses and their position on the sky image vary with respect to variations of the Sun`s position in the sky during a day and the geographical location of the PV modules within the PV array, the Ellipse Enclosed Cloud Coverage (EECC) also differs for different PV modules and hence, different irradiance values are obtained for different PV modules. The diffuse and ground-reflected irradiance components are assumed to be identical for all PV modules. In addition to the instantaneous SDIPs, daily irradiance time-series for different PV modules within the array are simulated by the model using a set of consecutive sky images. The results have shown that the obtained instantaneous irradiance values as well as irradiance time-series incident on different geographical locations within a PV array present a dispersed characteristic where the range of dispersion depends on the existing cloud type and its distribution in the sky. The model has been validated using different performance metrics and quite satisfactory validation results have been obtained verifying the model`s capability to generate reliable site-specific irradiance profiles or irradiance time-series to be utilized for different analysis purposes in PV arrays. The results of comparison of the measured and modeled Variability Index (VI) values, as a verification method of the performance of the developed model, show statistics of mean bias error (MBE) of 0.16, root mean square error (RMSE) of 2.394, correlation coefficient of 0.94 and mean absolute error (MAE) of 1.91. The proposed model is presented as a global model, with minimum dependency on sensors or other measurement equipment, which is capable of generation of SDIPs or irradiance time-series for any geographical location on the Earth`s surface under any type and distribution of cloud coverage in the sky, with available sufficient input data. A simple model is developed in MATLAB/Simulink environment in order to simulate PV modules based on one-diode mathematical model of a PV cell. PV arrays are simulated as combinations of series and parallel connected PV modules with different interconnection architectures and their performances are compared. A novel adaptive dynamic reconfiguration algorithm is developed in this thesis in order to improve power generation in PV arrays. The proposed reconfiguration algorithm is based on irradiance equalization method aiming at creation of series-connected rows of parallel-connected PV modules in a PV array with average irradiance values similar to the array`s average irradiance value in order to prevent limiting effect of less power generating rows on power generation by the array. The reconfiguration algorithm proposed in this thesis is a simple dynamic algorithm which intends to find near-optimal array configurations, in terms of irradiance equalization principle, based on the existing irradiance profiles as a result of the existing cloud coverage and distribution in the sky. The proposed algorithm is not limited to the number of PV modules included within a PV array and best fits large-scale centralized or distributed PV arrays rather than residential PV applications. The near-optimal array configurations are obtained by considering an irradiance threshold value which is in fact a tolerance shown against current limiting effects of the less power generating rows. The rows with average irradiance values falling within the irradiance tolerance from the array`s average irradiance value are not reconfigured by the algorithm since they do not cause significant limitations on array`s power generation. In this way, the number of reconfigured rows and PV module reconfigurations along with the number of required switching actions to obtain the final array configuration is reduced. Reduction of the number of switching actions also preserves lifetimes of switching devices. Switching actions are performed by a flexible switching matrix which is capable of connecting each PV module to each row of the PV array. The proposed reconfiguration algorithm is applied to totally four different irradiance profiles or partial shading scenarios and the array`s P-V characteristic curves are obtained before and after PV array reconfiguration considering different irradiance threshold values. The results have shown that the proposed reconfiguration algorithm is able to improve array`s power generation to almost maximum possible amounts under ideal case, in terms of irradiance equalization. In addition to achieving approximately maximum possible power generation in PV arrays, the results have also put forth the smoothing effects of the considered irradiance threshold value on array`s characteristic curves. Lower values of irradiance threshold have results in smoother characteristic curves to the cost of higher number of PV module reconfigurations and switching actions to be performed by the switching matrix. The results show that the reconfiguration algorithm has been able to improve array`s power generation by 4.7%, 6.1% and 2.7%, respectively under three different non-uniform irradiance profiles considered during analysis, where the reconfiguration algorithm is applicable. Although the proposed reconfiguration algorithm is applied to a single geographical location and time instant, it is expected to improve power generation in PV arrays at any geographical location or time interval. The proposed algorithm is also expected to perform better where the existing irradiance profiles contain irradiance values with high dispersion ranges. Keywords: Renewable Energy, Solar Energy, Photovoltaic, PV Module, Partial Shading, PV Power Generation, Spatially Dispersed Irradiance Profile, Irradiance Model, Cloud Cover, PV Array Reconfiguration, Switching Matrix, Adaptive PV System