An improved model-free predictive voltage control for grid-forming inverter with adaptive ultra-local data-model in renewable energy system

Conventional model-based predictive voltage control (MBPVC) for grid-forming inverters (GFIs) in renewable energy system is sensitive to parametric accuracy. To address this issue, an improved model-free predictive voltage control (MFPVC) is proposed for grid-forming inverter. First, the parametric impact on MBPVC is analyzed in GFI. Then, the adaptive ultra-local data-model (ULDM) of the GFI is established for model-free voltage prediction. The ULDM of GFI is updated in each control period by combining the capacitor voltage gradient relationship. The linear extended-state-observer with the adaptive strengthening factor is designed to enhance the performance of the ULDM. Additionally, the optimal switching sequence is proposed for further reducing voltage ripples. The duration of each voltage vector in corresponding optimal switching sequence is calculated based on the deadbeat principle. The proposed MFPVC method effectively eliminates parametric effect and improve the accuracy of model-free voltage prediction. Finally, the conventional MBPVC, conventional MFPVC and proposed MFPVC are compared by the designed hardware experimental platform of GFI.