CNN ‐based deep learning technique for improved H7 TLI with grid‐connected photovoltaic systems

Abstract

In this article, a three-phase transformerless inverter (TLI) for a solar photovoltaic (PV) system connected to a high-power grid are proposed, which has advantages of better performance and lower cost. The primary concern about the TLI is fluctuations in the common-mode voltage, which impacts switching frequency leakage current and grid interface system. An improved H7 common-mode voltage (CMV) clamped TLI with discontinuous pulse width modulation (DPWM) is designed using a conventional neural network (CNN)-based deep learning approach. In this, a completely minimized leakage current is obtained to avoid CMV transients. The proposed PV-connected improved H7-TLI provides low-loss DC-side decoupling, which further reduces leakage current and isolation of the PV system during off-grid. In addition, the effects of several factors on CNN deep learning performance are explored, including training data size, image resolution, and network configuration. The proposed technique has the potential to be used in a test instrument for intelligent signal analysis or used in an artificial intelligence system. Switching loss is analyzed using proposed and existing H7 inverters under different load conditions. To verify the theoretical explanation, existing H7 inverters is analyzed by MATLAB/Simulink, and the outcomes are tested experimentally. The total harmonic distortion (THD) analysis of proposed and existing topology is analyzed and compared. The THD values of the existing and proposed topology are 3.74% and 3.23%, respectively.