Solar Power Plant Optimization

1. Introduction

In the photovoltaic conversion chain, modules are responsible for converting energy from the sun into electric power, and several researchers attempted to improve the performance of the modules and their cells. This is a possible approach to optimize the performance of the photovoltaic system. However, in a photovoltaic system there are other devices that make up the conversion chain. Interaction among these stages should be considered in the system optimization, which gains fundamental importance when increasing its size. A problem associated with photovoltaic power plants is shading. There are some studies that discuss its impact on photovoltaic systems (Martínez-Moreno, Muñoz, & Lorenzo, 2010) (Sullivan, Awerbuch, & Latham, 2011) (Ubisse & Sebitosi, 2009). In these systems, the maximum power point tracking stage is very sensitive to shading, thus affecting the magnitude and morphology of the output curve (Petrone, Spanuolo, & Vitelli, 2007). To a large extent, this is due to the fact that Maximum Power Point Tracking (MPPT) algorithms are usually based on the assumption that the generated power curve has only one peak (Houssamo, Locment, & Sechilariu, 2010) (Moradi & Reisi, 2011) (Enrique, Durán, Sidrach de Cardona, & Andújar, 2010) (Esram & Chapman, 2007). Sanchez Reinoso, Milone, and Buitrago (2013) introduces the different aspects mentioned previously within the simulation, and the heterogeneity of static and variable cloud cover on the performance of the system is taken into account. The said paper does an exploratory analysis based on the simulation of the effects of shading, but it does not propose any methodology to optimize the system. The way of designing a scheme that makes the most of the energy received, in the case of highly-powerful plants under shading, is scarcely studied. On the other hand, it is known that there exist optimization problems that demand to explore a wide range of solutions, making the classic methods almost inapplicable.

In recent years, different approaches to deal with these problems have been proposed, for example, evolutionary algorithms (EA). Some interesting features from this technique are the simplicity of the operators used, the possibility of using fitness functions with very few formal requirements, and the ability to explore multiple points of the search space in every iteration (Sivanandam & Depa, 2008). In the photovoltaic area, different computational intelligence techniques were successfully applied on the detection of operation in island mode (Chao, Chiu, Li, & Chang, 2011), on radiation modeling (Koca, Oztop, Varol, & Koca, 2011) (Ozgoren, Bilgili, & Sahin, 2012) (Mellit, Mekki, Messai, & Kalogirou, 2011), and on simulators of autonomous systems (Mellit, Mekki, Messai, & Salhi, 2010b) (Gómez-Lorente, Triguero, Gil, & Espín Estrella, 2012). In Larbes, Ait Cheikh, Obeidi, and Zerguerras (2009) and Messai, Guessoum, and Kalogirou (2011), evolutionary algorithms are employed to optimize the maximum power point tracker controller, whereas in Mellit, Kalogirou, and Drif (2010a) the dimensioning of little isolated systems including batteries is optimized. In the case of grid-connected systems, the system optimization was also approached from the dimensioning point of view but using particle swarm optimization (Kornelakis & Marinakis, 2010). Other study that optimizes through evolutionary algorithms (Díaz-Dorado, Suárez-García, Carrillo, & Cidrás, 2011) focuses on the projection of shades among the installation components when using tracking mechanisms. It is noted in literature that although the effect shading has on photovoltaic systems has been proved, most optimization works focus on dimensioning (mainly, the array-inverter relationship), on static shades and on improving the maximum power point tracking algorithms. However, if the first stage is not optimized, the following stages, at best, would only be able to mitigate greater losses.