Application of hardware-in-the-loop for maximum power tracking of solar panels using artificial neural network

Abstract

Hardware-in-the-Loop (HIL) is a technology for simulating hardware in a highly reliable iterative manner without the need to create a physical system, which reduces costs and shortens system development time. This paper presents the application of the HIL404 model for the maximum power point tracking (MPPT) system of solar panels using an artificial neural network (ANN) method. The system includes solar panels connected to a boost converter, which adjusts the optimal duty cycle to supply maximum power to the load. The ANN model, created in Matlab/Simulink through the fmi function, simplifies the simulation process. Data on solar radiation, temperature, and voltage were collected from seven DXP72-Z-330 solar panels with a voltage rating of 261 V at a maximum power of 2,300 W over one week to train the feedforward ANN using backpropagation. The training dataset consists of solar radiation intensity and temperature, with the target output being the system voltage that maximizes power. The paper evaluates the optimal transfer function for single-layer and two-layer hidden neural networks. Results indicate that the Linear function is most effective, with a mean absolute percentage error of 31.114 for the single-layer model and 8.248 for the two-layer model, showing the two-layer network to be more accurate. Additionally, the paper compares the performance of the MPPT system using both ANN models with the perturb and observe (P&O) method on a dataset of 1,200 data points. The load is a 45-ohm resistor. The test results show that the two-layer hidden neural network's MPPT system produced slightly higher power at the load than the P&O method and was able to accurately track the maximum power.