Neural Network Based Active Power Curtailment at Converter Stage for Voltage Rise Mitigation in High Photovoltaic Penetrated Low Voltage Distribution Networks
Journal of Applied Sciences and Environmental Management · 2026 · 30 (6) · 1901-1907
DOI: 10.4314/jasem.v30i6.20 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Applied Sciences and Environmental Management, 2026. Verfasst von M. Aminu und 3 weiteren Autor:innen.
Abstract (Originalsprache)
The increasing penetration of rooftop photovoltaic (PV) systems in low voltage (LV) distribution networks causes voltage rise during high generation and low load periods, which may violate the EN 50160 limit of 1.10 p.u. and trigger protection disconnection. This paper proposes a neural network‑based active power curtailment (ANN‑APC) at the converter stage for voltage rise mitigation in high photovoltaic (PV) penetrated low voltage (LV) distribution networks using a single control parameter (duty cycle) instead of the conventional inverter‑stage approach that requires six switching pulses. A 3‑bus LV distribution network (33 kV, Nigerian specifications) with two 100 kW grid‑connected PV systems is modelled in MATLAB/Simulink. An ANN with 6 inputs (irradiance, temperature, bus voltages, loads), 10 hidden neurons, and 2 outputs is trained via back‑propagation on 1000 simulation scenarios (70% training, 30% testing). The trained ANN predicts the required duty cycle to curtail active power when voltage exceeds the limit. Results show that the ANN achieves a final MSE of 6.2×10⁻⁵ and R=0.998. Without control, voltage at the line end (Bus 3) reaches 1.085 p.u. at 1000 W/m². With ANN‑APC, all bus voltages remain ≤ 1.10 p.u. (99.8% compliance) with only 15.6 kW (15.5%) curtailment. Compared to conventional inverter‑stage APC, the proposed method reduces response time by 58% (0.05 s vs. 0.12 s), settling time by 49% (0.18 s vs. 0.35 s), overshoot by 61% (0.7% vs. 1.8%), and steady‑state error by 50% (±0.004 p.u. vs. ±0.008 p.u.). Compared to fixed droop control, curtailed energy is reduced by 33.3% (12.4 MWh/year vs. 18.6 MWh/year). The proposed ANN‑APC method offers a simple, fast, adaptive, and cost‑effective solution for voltage rise mitigation, particularly suitable for retrofitting existing PV systems in developing countries.
Zitieren
APA
Aminu, M., Hammanadama, Y. S., Musa, S. Y., & Usman, R. (2026). Neural Network Based Active Power Curtailment at Converter Stage for Voltage Rise Mitigation in High Photovoltaic Penetrated Low Voltage Distribution Networks. Journal of Applied Sciences and Environmental Management, 30(6), 1901-1907. https://doi.org/10.4314/jasem.v30i6.20
Vancouver
Aminu M, Hammanadama YS, Musa SY, Usman R. Neural Network Based Active Power Curtailment at Converter Stage for Voltage Rise Mitigation in High Photovoltaic Penetrated Low Voltage Distribution Networks. Journal of Applied Sciences and Environmental Management. 2026;30(6):1901-1907. doi:10.4314/jasem.v30i6.20
BibTeX
@article{Aminu2026,
author = {Aminu, M. and Hammanadama, Y. S. and Musa, S. Y. and Usman, R.},
title = {Neural Network Based Active Power Curtailment at Converter Stage for Voltage Rise Mitigation in High Photovoltaic Penetrated Low Voltage Distribution Networks},
journal = {Journal of Applied Sciences and Environmental Management},
year = {2026},
volume = {30},
number = {6},
pages = {1901--1907},
doi = {10.4314/jasem.v30i6.20}
}
EUVIMED