Data-driven Infrastructure Planning for Offshore Wind Farms

近年来，英国在海上风能中的份额一直在稳步增长；目前海上风电装机容量为10.4吉瓦（参考：英国风能数据库在线：https://www.renewableuk.com/page/UKWEDhome/Wind-Energy-Statistics.htm）。已经有计划开发容量超过1千兆瓦的大型海上风力发电场。然而，与传统发电技术相比，如何最好地开发、维护和运营风力发电场及其基础连接系统，以实现成本竞争力，仍存在一定程度的不确定性。例如，位于离岸较远的大型涡轮机更难接近。此外，海上资产的可及性取决于天气条件，这可能对收入和支出产生重大影响。因此，作业者并不总是拥有足够的信息来做出与资产规划和维护相关的最具成本效益的决策。类似的问题也出现在海上风电资产投资和整合的长期决策的不确定性中。现有的建模方法简化了假设，不幸的是导致了次优解决方案，特别是在许多因素都在起作用的大型风力发电场。例如，现有工具通常使用两种状态的马尔可夫链进行故障和修复，这意味着故障和修复时间是指数分布的。虽然这可能是对故障的合理近似，但由于有时根据天气条件到达资产的时间很长且随机，因此维修很少呈指数分布。Durham在更准确地表示故障和修复过程以及处理建模假设方面具有经验，特别是在规划阶段的严重不确定性下。该项目的目的是为大型海上风电场的规划和运营开发建模和优化决策的新方法，特别是在面临可用可操作信息不确定的情况下。我们还旨在将故障和维修与风电场的环境条件和单个涡轮机的运行条件联系起来。我们希望这能提高故障和维修的预测能力，包括考虑涡轮机的可及性，从而降低成本。该项目的结果将为未来几年的决策过程提供信息，以便更有效地整合和运营大型海上风电场，为未来以更稳健和更具成本效益的方式发展铺平道路。

The UK's share of offshore wind energy has been steadily increasing in recent years; there is now 10.4 GW of installed capacity offshore (reference: UK Wind Energy Database online: (https://www.renewableuk.com/page/UKWEDhome/Wind-Energy-Statistics.htm). There are already plans for developing large-scale offshore wind farms with capacities exceeding 1 GW. There remains however a degree of uncertainty over how to best develop, maintain and operate the wind farms and their underlying connection system to achieve cost competitiveness compared to conventional generation technologies. For example, larger turbines located farther offshore are more difficult to access. Further, accessibility of offshore assets depends on weather conditions, which can have a significant impact on income and expenditure. Consequently, the operator does not always possess sufficient information to make the most cost-effective decisions relating to planning and maintaining their assets. Similar problems arise from uncertainty over long-term decisions for investment and integration of offshore wind assets. Existing modelling methods make simplifying assumptions which unfortunately lead to suboptimal solutions, especially in larger wind farms where many factors are at play. For instance, existing tools typically use two state Markov chains for failure and repair, meaning that failure and repair times are exponentially distributed. Whilst this can be a reasonable approximation for failure, repair is rarely exponentially distributed due to sometimes large and random lead times to reach the assets depending on weather conditions. Durham has experience with more accurately representing failure and repair processes and handling modelling assumptions, especially under severe uncertainty in the planning stage. The aim of this project is to develop new methods for modelling and optimising decisions involving planning and operation for larger offshore wind farms, especially when facing uncertainties in the available actionable information. We also aim to link failure and repair to the environmental conditions of the wind farm, and operational conditions of individual turbines. We expect this to improve the predictive capability of failures and repairs, including the consideration of turbine accessibility, thereby reducing costs. The outcome of this project will inform decision-making processes for years to come for efficient integration and operation of larger offshore wind farms paving the way for future developments in a more robust and cost-effective manner.