Wind Farm Operations and Maintenance Scheduling Collaborative Optimization

風電場運行與維護調度的協同優化研究

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Related Research Unit(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date25 Jul 2022

Abstract

In recent years, a surge of wind power installations has driven the rapid growth of wind farms. Wind energy plays an essential role in accelerating the global energy transition and achieving carbon neutrality. However, the high uncertainty and randomness of wind power generation bring significant challenges to both wind farm operations and the integration into the power grid. Besides, as wind farms are usually located in remote sites with harsh environmental conditions, wind turbine maintenance is also an indispensable element in wind farm operation management. Thus, the wind farm O&M problem is an important topic and has been studied by developing various techniques, such as efficient strategies and optimization methods.

In the face of complicated internal and external operating conditions, the coupling characteristic of interactions between wind turbine operations and maintenance activities, as well as distinctive operational objectives of different stakeholders in the operation process, the wind farm O&M optimization problem still needs thorough research. To solve the above issues, four research works are conducted in this thesis, which are presented as follows.

Firstly, focusing on a single offshore wind farm, the optimization problem of scheduling both maintenance tasks and power productions in consideration of their distinct scheduling timescales and the wind power uncertainty is studied. A two-stage robust optimization model is formulated to arrange wind turbine maintenance tasks and maximize wind farm power productions. The proposed model is equivalently transformed to a form which can be efficiently solved by decomposition algorithms. Computational results show that the developed model provides reliable maintenance schedules that are robust against wind power uncertainty and have minimal impact on wind farm power production.

Secondly, extending the scope of the first study, the collaborative O&M optimization problem of multiple offshore wind farms operating in a one-to-many maintenance service mode is studied. A decentralized robust optimization model is developed, in which wind farms and a maintenance service provider are treated as individual stakeholders with different interests but coupled by maintenance resources, such as vessels and technicians. Based on the analytical target cascading algorithm, the model is decoupled into independent models for each stakeholder and can be easily solved. Computational results illustrate that the developed model can coordinate maintenance resources among multiple wind farms and generate individual robust O&M schedules for them without sharing much private operational information.

Next, to further tackle risks induced by the wind power uncertainty and promote the wind farm integration, a bi-level robust approach for optimizing multi-timescale operations of a wind farm is studied. The day-ahead power supply to the grid is planned via a two-stage robust optimization model. Subsequently, intra-day wind farm operational instructions are fine-tuned via a multi-stage robust optimization model using updated operational information. The proposed models are solved by using computationally tractable optimization methods. Computational results verify that the proposed approach can guide wind farm operators to make more reliable power supply plans to the grid while providing robust operation schedules with higher flexibility and less conservativeness.

Finally, to effectively utilize real operational data and provide general maintenance schedules, a data-driven robust method for generating the optimal wind farm maintenance scheduling template under various wind power uncertainty is studied. A distributionally robust optimization model with a φ-divergence based ambiguity set is formulated to schedule maintenance tasks while optimizing the expected scheduling performance in the worst-case wind power distribution. The proposed model is easily solved after the equivalent transformation into its computationally tractable formulation. The optimal wind power ambiguity set is chosen by learning techniques using real data. Computational results demonstrate that the proposed method can generate robust maintenance schedules with better performance and less conservativeness, even with a small size of available data.