Search documents

Browse topics

Document details

Data-Driven PV System Engineering: Comprehensive Modelling and Simulation of a Grid-Connected Three-Phase Inverter
V. Titova, L. Lahrmann, A.A. Alves, F. Haase, Z. Khadiri-Yazami, M. Lapke
PV System, Simulation, Inverter, Modelling
PV Systems Engineering, Integrated/Applied PV
Subtopic: Power Electronics and Electrical Grid Interface
Event: 8th World Conference on Photovoltaic Energy Conversion
Session: 4EO.2.5
1161 - 1164
ISBN: 3-936338-86-8
Paper DOI: 10.4229/WCPEC-82022-4EO.2.5
0,00 EUR
Document(s): paper


In a data-driven modelling approach, the intelligent integration of data between physical and virtual entities of a photovoltaic (PV) system can facilitate innovation, control and optimization of the entire PV power plant. In an inverter, the insulated-gate bipolar transistor (IGBT) power electronics are used to distribute and convert the direct current from the PV module to alternating current. Due to the demand for performance improvements, the need to model these IGBT stacks is constantly increasing [1, 2]. Consequently, to reduce thermal resistance and improve the reliability of power electronics, new data points such as thermal loading and internal humidity are increasingly moving into the scientific focus [3]. In this contribution, we pursue two approaches towards anomaly detection in the operation of a PV plant. A comparison-based approach, analyzing the operation-induced data of an inverter to a physically motivated inverter model, and a failure-motivated approach using numerical modelling methods to reveal the hidden characteristics for latter data analysis. We present a MATLAB/Simulink model of a three-phase grid-connected inverter. We demonstrate the responsiveness of our model by changing the polarity in the active and reactive current control. This model shall serve as an idealized digital twin of the inverter to aid in anomaly detection within measured inverter data. Furthermore, we present initial ideas on a failure-motivated modelling approach to improve our model’s accuracy and enable optimized predictive maintenance algorithms. The use of hidden characteristics, such as IGBT junction temperature and power module interior humidity can serve as markers for failure detection in the operation of a PV system. Therefore, we have analyzed initial power loss calculations of an IGBT, which serve as a basis for thermal loss simulations to predict overheating of the switch components within the inverter. We examine the correlation between modelled and measured temperatures of the IGBT stacks. In addition, we propose a humidity model to prevent humidity-induced failures in power electronics.