A New Nonsmooth Optimal Control Framework for Wind Turbine Power Systems

TL;DR

This paper introduces a novel nonsmooth optimal control framework for wind turbine power systems using a nonlinear DAE model, enabling better control during operational transitions and across all wind speeds.

Contribution

The paper develops the first nonsmooth optimal control framework for WTPS using a nonlinear DAE model, addressing limitations of previous linearized models and smoothing assumptions.

Findings

Effective control during operational transitions

Robust performance with real-world wind data

Superior results compared to smoothing and naive methods

Abstract

Optimal control theory extending from the calculus of variations has not been used to study the wind turbine power system (WTPS) control problem, which aims at achieving two targets: (i) maximizing power generation in lower wind speed conditions; and (ii) maintaining the output power at the rated level in high wind speed conditions. A lack of an optimal control framework for the WTPS (i.e., no access to actual optimal control trajectories) reduces optimal control design potential and prevents competing control methods of WTPSs to have a reference control solution for comparison. In fact, the WTPS control literature often relies on reduced and linearized models of WTPSs, and avoids the nonsmoothness present in the system during transitions between different conditions of operation. In this paper, we introduce a novel optimal control framework for the WTPS control problem. We use in our formulation a recent accurate, nonlinear differential-algebraic equation (DAE) model of WTPSs, which we then generalize over all wind speed ranges using non-smooth functions. We also use developments in nonsmooth optimal control theory to take into account nonsmoothness present in the system. We implement this new WTPS optimal control approach to solve the problem numerically, including (i) different wind speed profiles for testing the system response; (ii) real-world wind data; and (iii) a comparison with smoothing and naive approaches. Results show the effectiveness of the proposed approach.