Artificial Intelligence Based Control Strategy of a Three-Phase Neutral-Point Clamped Back-to-Back Power Converter with Ensured Power Quality for WECS

• Main Authors: Uddin, M. Nasir, Tabrizi, Yazdan H.
• Format: Conference Proceeding
• Language: English
• Subjects: back-to-back power converter; Bandwidth; Costs; Machine learning algorithms; Machine learning approach; neutral-point-clamped inverter; PI control; Power quality; Total harmonic distortion; Training; wind energy conversion system and total harmonic distortion
• ISSN: 2576-702X
• DOI: 10.1109/IAS54023.2022.9939815

This paper provides an in-depth investigation into the state-of-the-art power converter control approaches utilizing artificial intelligence that ensure the power quality for wind energy conversion systems (WECS). The most promising and feasible wind energy conversion configuration has been elected to be evaluated in an attempt to reduce the computing cost and time, as well as meet the grid code requirements. For this purpose, in this work, a back-to-back neutral-point clamped power converter is controlled with high precision using a machine learning algorithm. The machine is trained offline by data acquired from the wellknown voltage-oriented control (VOC) technique. The majority of the computational load is moved from online to offline mode. Thus, there is no need for accurate development of the PI controller parameters and bandwidth, and hence, the cost and time of calculation will be considerably reduced. As a consequence, the recommended machine learning-based technique can take over the conventional VOC responsibilities. To accomplish this, the training dataset is applied to learn the behavior of the system using a locally weighted lasso regression approach. The cost function is then minimized using stochastic gradient descent, batch gradient descent, Broyden-Fletcher-Goldfarb-Shanno (BFGS), and limited-memory BFGS optimizers, successively. The comparative analysis reveals that the BFGS family of optimizers outperforms the counterparts in terms of computation time, accuracy, and THD performance of WECS.