Andrej Horvat

Mr. Andrej Horvat

orcid

Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
30 September 2026 in Advanced Engineering Letters
Verified
Journal Article
Open access
Damir BlaževićDamir Blažević
Damir Blažević
Tomislav KeserTomislav Keser
Tomislav Keser
AHAndrej Horvat
Andrej Horvat
DNDamir - Nozica
Damir - Nozica
This paper compares a single-diode model (SDM) parameter extraction method based on differential evolution (DE) with the analytical Villalva method, using a selected set of current-voltage (I-V) curves from the publicly available Sandia flash test dataset. The paper examines the methods’ applicability to experimentally measured curves from commercial photovoltaic (PV) modules. The SDM DE method analytically eliminates parameters from five to three, uses a Newton-Raphson solver for the implicit I-V characteristic, and minimises an objective function based on the error at the maximum power point. The results show that the proposed method achieves a lower root mean square error (RMSE) and normalized root mean square error (NRMSE) than the Villalva method across all tested curves, though at a higher computational cost.

5 October 2026 in Applied Sciences
Verified
Journal Article
Open access
PDF file
Tomislav KeserTomislav Keser
Tomislav Keser
DNDamir - Nozica
Damir - Nozica
Damir BlaževićDamir Blažević
Damir Blažević
AHAndrej Horvat
Andrej Horvat
When two photovoltaic (PV) strings are connected in parallel to a single maximum power point tracker (MPPT) input, the tracker enforces one operating voltage on both strings, and whichever string is displaced from its own maximum power point contributes a power deficit. This loss is distinct from, and additive to, the mismatch loss arising among series-connected modules within each string. This work quantifies the two contributions separately and proposes a genetic algorithm (GA) that assigns modules from a real manufacturer flash list to strings and MPPT inputs so as to maximize the power delivered at the tracker, rather than the power of each string to be considered in isolation. Each module is represented by a double-diode model with parameters extracted from its flash list entry using a hybrid differential evolution procedure. String characteristics are obtained by summing module voltages at common current, and MPPT characteristics are obtained by summing string current at common voltage, so that the objective function evaluates the true interconnected operating point. For 68 modules of a 43.5 kW array, the mismatch loss is 0.398 W with separate MPPT inputs and 1.534 W with two strings per input. In the first case, sorting by MPP current lies within 0.122 W of the GA; in the second, all six criteria fall within 1.069 W of one another, random grouping ranks second, and the GA gains 3.243 W over the best of them. Sorting by a single parameter thus redistributes the loss between levels rather than removing it.

14 October 2026 in Eng
Verified
Journal Article
Open access
PDF file
Damir BlaževićDamir Blažević
Damir Blažević
Tomislav KeserTomislav Keser
Tomislav Keser
DNDamir - Nozica
Damir - Nozica
AHAndrej Horvat
Andrej Horvat
This paper represents the expansion of the parameter extraction method for a photovoltaic module based on differential evolution (DE) developed for the double-diode model (DDM). DDM takes into account two diodes that model the recombination losses in the depleted layer, which means that the model has seven unknown parameters. By fixing the diode ideality factors and analytically eliminating the saturation current of the first diode and the photocurrent from the boundary conditions defined by the data given from the manufacturer’s datasheet, the dimensionality of the DE search comes down from seven to three parameters. The proposed approach is compared and validated using a single well-known multicrystalline KC200GT module. The results show that the proposed method achieves RMSE of 0.078627 A and NRMSE of 1.1520%, while Moshksar’s method achieves RMSE of 0.078783 A and NRMSE of 1.1543% for standard test conditions.