Time-delayed pith angle control of wind turbine systems-based Smith ultralocal model machine learning technique

Abstract The time-varying nonlinearity and uncertain nature of the wind turbine (WT) systems lead to changes of an operating point, which necessitates a powerful control methodology to regulate such complex systems. The challenges of regulation of WT systems have been intensified in the presence of time delay. In this research, the pitch angle regulation of a 2-mass WT model with time delay is addressed by the ultralocal model (ULM) control scheme, by means of Smith predictor (SP). In particular, to deal with the difficulties in adjusting the ULM coefficients, we adopt a deep deterministic policy gradient (DDPG) as the tuner mechanism of these coefficients. Firstly, we formulate the ULM scheme in a model-independent manner for the pitch angle control of WT to ameliorate the plant dynamic performance. Secondly, we incorporate the SP into the control structure to compensate for the effects of time delay in the WT plant. Thirdly, we employ the DDPG with the actor-critic framework for adaptive autotuning of the ULM controller coefficients. The analysis and tuning are of tutorial value for practitioners and engineers, and the usefulness and applicability are verified by a comparative simulation.