Solar Irradiation Changes Detection for Photovoltaic Systems Through ANN Trained with a Metaheuristic Algorithm Chapter in Scopus uri icon

abstract

  • © 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.Nowadays, Photovoltaic (PV) energy sources are responsible for more than 720TWh of the global energy consumed. Therefore, among all the arising research topics regarding PV sources, Maximum Power Point tracking (MPPT) algorithms are widely studied, since they have critical importance in order to properly track and acquire the Maximum Power Point (MPP) of energy that can be harvested from the PV source; where, metaheuristic optimization algorithms have been adapted as MPPT solutions, showing an improved settling time with steady-state oscillations reduction. Yet, many of the Metaheuristic-based MPPTs have issues against dynamically variable MPP through time, since they cannot work properly without a reset signal upon temperature and irradiance parametric changes, due the fact that those algorithms converge on a global solution after some iterations and need to know when to start looking for the MPP again in order to avoid getting stuck in old MPP points. Hence, this work shows the implementation of Artificial Neural Networks (ANN) for pattern recognition through the power data acquisition, which enables an efficient solution that allows detecting changes in solar irradiation, which leads to a reliable reinitialization signal through the detection of solar irradiation through the ANN; where, results show over 99% of accuracy from the change in solar irradiation detection. Moreover, the ANN was trained using the metaheuristic Earthquake Algorithm (EA), which was again validated as a training method for ANN. In addition, this work enables other users to suite the EA algorithm to other applications and the implementation of ANN into the Simulink environments; both achievements, through the presentation and exploration of a complete MATLAB code example for the implementation of the EA as ANN training method, and the complete Simulink testbed design with a full description of the ANN implementation for the Simulink environment.

publication date

  • January 1, 2021