Optimization of PV Systems Using Linear Interactions Regression MPPT Techniques

Supervised machine learning techniques such as artificial neuro-fuzzy inference system (ANFIS) and artificial neural network (ANN) are powerful techniques used to extract maximum power from photovoltaic systems. However, these offline methods require large and accurate training datasets for effective MPPT. This paper presents an advanced use of the linear regression with interactions (LIR) technique that can produce large and very accurate training datasets needed for MPPT improvement. To confirm the success of the LIR technique, combination of LIR and ANFIS as LIR-ANFIS technique results was compared with conventional ANFIS results, and that of bootstrap aggregation (bagged) and boosted tree ensemble regression as bagged-ANFIS and boosted-ANFIS results under different weather conditions. Results show that LIR-ANFIS technique yielded the best result and with improved performance.