Daria Madjidian

A distributed power coordination scheme for fatigue load reduction in wind farms

We consider a scenario where a wind farm is given a power set point below its actual power production capacity. The objective is to dynamically redistribute power in order to minimize the fatigue loads experienced by the turbines, while maintaining the desired power production at all times. We show that this can be done in a distributed way by coordinating neighboring turbines. The result is a control scheme where both the synthesis and the resulting control law only require each turbine to communicate with a limited set of neighboring turbines.

Distributed low-complexity controller for wind power plant in derated operation

We consider a wind power plant of megawatt wind turbines operating in derated mode. When operating in this mode, the wind power plant controller is free to distribute power set-points to the individual turbines, as long as the total power demand is met. In this work, we design a controller that exploits this freedom to reduce the fatigue on the turbines in the wind power plant. We show that the controller can be designed in a decentralized manner, such that each wind turbine is equipped with a local low-complexity controller relying only on few measurements and little communication. As a basis for the controller design, a linear wind turbine model is constructed and verified in an operational wind power plant of megawatt turbines. Due to limitations of the wind power plant available for tests, it is not possible to implement the developed controller; instead the final distributed controller is evaluated via simulations using an industrial wind turbine model. The simulations consistently show fatigue reductions in the magnitude of 15-20%.

A Stationary Turbine Interaction Model for Control of Wind Farms

Turbines operating in wind farms are coupled by the wind flow. This coupling results in limited power production and increased fatigue loads on turbines operating in the wake of other turbines. To operate wind farms cost effectively, it is important to understand and address these effects. In this paper, we derive a stationary model for turbine interaction. The model has a simple intuitive structure, and the parameters a clear interpretation. Moreover, the effect of upwind turbines on a downwind turbine can be completely determined through information from its closest neighbor. This makes the model well suited for distributed control. In an example, we increase total power production in a farm, by coordinating the individual power production of the turbines. The example points to an interesting model property: decreasing power in an upwind turbine can lead to a larger and larger increase in available power downwind. (Less)

On Using Wind Speed Preview to Reduce Wind Turbine Tower Oscillations

We investigate the potential of using previewed wind speed measurements for damping wind turbine fore-aft tower oscillations. Using recent results on continuous-time H2 preview control, we develop a numerically efficient framework for the feedforward controller synthesis. One of the major benefits of the proposed framework is that it allows us to account for measurement distortion. This results in a controller that is tailored to the quality of the previewed data. A simple yet meaningful parametric model of the measurement distortion is proposed and used to analyze the effects of distortion characteristics on the achievable performance and on the required length of preview. We demonstrate the importance of accounting for the distortion in the controller synthesis and quantify the potential benefits of using previewed information by means of simulations based on real-world turbine data.

Distributed Control with Low-Rank Coordination

A common approach to distributed control design is to impose sparsity constraints on the controller structure. Such constraints, however, may greatly complicate the control design procedure. This paper puts forward an alternative structure, which is not sparse yet might nevertheless be well suited for distributed control purposes. The structure appears as the optimal solution to a class of coordination problems arising in multiagent applications. The controller comprises a diagonal (decentralized) part, complemented by a rank-one coordination term. Although this term relies on information about all subsystems, its implementation only requires a simple averaging operation.

Decentralized feedforward control of wind farms: Prospects and open problems

The problem of load mitigation in wind turbines located in wind farms is addressed. The benefits of letting turbines communicate with and account for their neighbors are explored. First, the idea of exploiting previewed wind speed measurements is examined. The problem is formulated as an H2 model matching optimization. The influence of preview length on the performance is analyzed and simulation results are presented. Then, the possibility of cooperation between turbines is studied within a distributed feedforward control scheme. The problem is formulated as a decentralized model matching optimization and several theoretical challenges associated with it are outlined. An approximate frequency domain solution and preliminary simulation results are presented.

A Decomposition Result in Linear-Quadratic Coordinated Control

Scalability is a fundamental requirement in control design for large-scale systems. Typically, it needs to be considered explicitly at the expense of performance degradation and more complicated design procedures. In this paper, we present a class of large scale systems where scalability is an inherent property of the optimal centralized solution. More specifically, we study a coordination problem where a group of identical subsystems are required to satisfy an equality constraint on the sum of their inputs. We show that the problem can be completely decomposed in terms of the unconstrained problems associated with each subsystem. In particular, the computational effort required to obtain the optimal solution is independent of the number of subsystems and the only global information processing required to execute the optimal control law is a simple summation, which scales well when the number of subsystems grows large.