• Overview
• Visitor Info
• People

• Publications

• Courses
• GA/SA/MA

• News
• Seminars

Scenario-Based Model Predictive Control for Wind Turbines

Wind Energy

Increasing energy consumption poses a pressing worldwide challenge, with CO2 emissions causing climate change and the dependence on fossil fuels threatening economic sustainability. Given current global developments, these issues will become even more serious over the coming decades. For example, the International Energy Agency (IEA) predicts that the global greenhouse gas emissions by the energy sector will increase by 130% by 2050. While a portfolio of actions is required to mitigate this development, improvements in the technology of renewable energy sources is a key element of an effective strategy.

While wind power still accounts for a fairly small share of the global energy mix (ca. 1% worldwide and 2% in OECD countries), the penetration is already much higher in some countries (e.g. Germany 4%, Spain 11%, Portugal 9%, Denmark 20%), according to IEA figures. Moreover, it is one of the world's fastest growing renewable energy sources, with an average yearly growth of almost 30% of the installed capacity since 2000. For example, in 2008 the global investments for new installations amounted to more than $45 billion in 2008, corresponding to over 50% of global investments in renewables (IEA figures). Indeed, while currently behind hydropower and ahead of biomass, at the current growth rates wind power is projected to soon become the most significant source of renewable-based electricity.

Onshore wind turbine technology is technologically proven and economically competitive in areas where the resource is strong (costs per MWh range between 70-130 USD). Technological development, however, can help in further reducing the costs of wind energy (together with economies of scale), and may also enhance its competitiveness by extending the operating regions of wind turbines (e.g. offshore, etc.).

The main technological challenges for wind power are to reduce the initial infrastructure cost and the maintenance costs, to increase the variability of the resource, and to address the low density of generated power per unit area. These issues motivate the significant research and development efforts that are being carried out all over the world to create larger wind turbines, with relatively lighter components, and to reduce the fatigue effects of vibrations due to wind turbulences. At the same time, regulations for wind power plants are becoming more strict in terms of amount and quality of the generated electricity.

Project Description

In the above context, active control systems can play a crucial role in increasing the power capture, extend the lifetime, and improve the reliability of wind turbines. The design of control systems for wind turbines, able to produce a high power output while limiting the loads on the turbine (thereby reducing maintenance costs and material fatigue), is made difficult by the presence of input saturation and stochastic nature of the wind.

Recently, researchers have proposed to apply Model Predictive Control (MPC) because of its strengths in handling multi-input multi-output (MIMO) systems and input constraints, and because of its close relationship to optimal control. Various simulation results suggest that a significant improvement is possible from conventional state-of-the-art PI controllers. However, the stochastic uncertainty of the wind speed remains a major challenge.

The goal of this project is to apply a new approach called Scenario-Based MPC (SMPC) to wind turbines. Its basic idea is to create multiple scenarios of the uncertain wind speed and compute the optimal control solution that works robustly under each scenario. Besides its highly intuitive concept, the approach has several practical advantages: it works for any type of probability distributions (in fact, the distribution may be unknown), it is very easy to apply, the handling of bi-linear terms in the disturbance and state/inputs is straightforward, and it is computationally very efficient. Previous studies have already found SMPC to outperform other control strategies in similar settings (e.g. buildings control).

The major tasks/requirements of this project are as follows:

• Acquisition of some background on MPC and scenario-based optimization
• Installation of the FAST wind turbine model (openly available, with Simulink interface)
• Derive a simplified turbine model for MPC (by first principles or system identification)
• Create a stochastic model for the wind speeds
• Setup the SMPC problem, by formulating a sensible cost function and constraints
• Simulation of the SMPC in closed-loop operation, comparison to other baseline controllers