Estratégia de diagnóstico inteligente de falhas para sistemas fotovoltaicos de alta potência

Abstract

In the last years, the instalation of photovoltaic systems had a exponecial growth, duo to government incentives, concerns with the environment e also a decrease in prices of photovoltaic modules. This increase also drives a greater number of commissioning, monitoring and maintenance required in these systems. This cited procedures are necessary to ensure a greater efficiency of the photovoltaic systems, and consequently, a greater generation. However, the photovoltaic modules are installed outdoors, they are prone to damage caused by big temperature variation, humidity, heavy wind and hailstorm. This conditions, when extreme, can cause damage on the photovoltaic module that entail in huge losses and even fires. Furthermore, these faults that occur on the contionuos current side are harded to harder to recognize by the existing equipment on the market, for having different characteristics of the faults that occur on the alternating current side. Therefore, this dissertation proposes a complete system to diagnose faults in photvoltaic arrays based on the current per voltage curve and artificial neural networks. This system is composed by the array to be analysed, the DC-DC converter, which is responsible to trace the I-V curve experimentally, the command circuit, controls the converter, and also, the data acquisition system, responsible for store the curve. The results of the proposed system shows that the DC-DC flyback converter is capable of tracing the I-V curve for all failure conditions of the photovoltaic array. I addition, on the artificial neural network results was observed that the network is able to diagnosing the proposed conditions with 99,3% of accuracy, classifing correctly the faults. Experimental tests showed that the modular flyback converter was able to divide the voltage stresses applied to its input, and is also able to trace the I-V curve of the photovoltaic module.