DESIGN AND PERFORMANCE ANALYSIS OF ARTIFICIAL INTELLIGENCE BASED MAXIMUM POWER POINT TRACKING ALGORITHMS FOR PV ARRAY UNDER PARTIAL SHADING CONDITIONS

Abstract

Sustainability is the ability of a resource to be viable between generations and not get exhausted or affect the nature directly or indirectly. The rapidly increasing demand and the inability of the conventionally utilized fossil fuels does not add to the sustainability of the environment. It not only pollutes the environment via its fumes but also leaves a large carbon footprint. Renewable energy sources are the sources which do not exhaust over time. Solar energy is a widely used source of energy due to the abundant availability of sunlight all over the world.

The main objective of any generation activity deals with the maximization of benefits and revenue. This revenue depends upon the generation by the plant, which in turn depends on the plant performance. Thus, a maximization of the plant performance i.e., the output power is desirable. This maximization of power can be obtained from a PV array by operating it at its maximum power point i.e., the MPP of the designed array.

Various MPP techniques are employed for this purpose and they vary from each other on the basis of their inherent structure, the logic behind the algorithm, the convergence as well as computation time. Uniformity of irradiance is a rare phenomenon in the field of solar power generation. However, testing of the PV arrays is done at STC i.e., 1000 W/m2 irradiance at a temperature of 25 ̊C or 298 K. Conventionally available techniques have been used over years. However, the inherent oscillations in their response and the relatively poor tracking speeds have acted as a motivation for the development of Artificial Intelligence (AI) based techniques.

The AI based techniques employ the basic input required by the conventional techniques along with an additional change in error of these inputs. This gives these techniques an effective direction for locating and approaching the MPP at a faster speed within least possible time. The three main AI based controllers are Fuzzy Logic Controller (FLC), Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference Systems (ANFIS). Rigorous comparative studies are performed on the controllers and a continuous effort is put in to further boost their performance as compared to their conventionally available counterparts by incorporating one modification or the other. A simplified and optimized FLC (SOFLC) is proposed, designed and tested to approximate the performance of a 49-rule base FLC which acts as a trade-off between complexity and accuracy by merely 4 rules and a compensating polynomial.