The grid integration of offshore wind energy systems can have a sigificant impact on offshore wind energy system performance. Multiterminal HVDC networks (grids) bring significant benefits, however, effective protection system design is essential to ensure that faults on the HVDC system do not result in transient disconnection of offshore wind farms. While recent work has demonstrated effective performance for novel HVDC circuit breaker topologies, their exact configuration on a grid is still unknown. The placement of the HVDC circuit breaker can lead to significant differences in the offshore grid impact. Even with significant investment in HVDC circuit breakers, there may still be a requirement for long duration curtailment of offshore wind.

Selection of the HVDC grid configuration, and the protection of the HVDC grid, can have a significant impact on the continous operation of connected offshore wind farms. In the worst case, a fault on the HVDC system can necessitate large scale curtailment of the connected offshore wind. Without adding redundant cables (or reducing power flows to allow for redundancy), extensive HVDC circuit breaker application will still not fully mitigate the risk of wind farm disconnection on all grids. This presentation will provide a summary of the trade-offs of HVDC protection system design for integration of offshore wind, providing an outlook for effective protection system design and the associated studies that should be performed.

Functional testing is essential to validate the performance of offshore wind assets, and work is underway to demonstrate effective system integration with offshore wind energy systems and HVDC grids. The presentation will comment on likely means of functional testing for different design, commissioning and testing stages within the project lifecycle.

Some of the best offshore wind resources are located in areas where the onshore grid is weak and may lack sufficient capacity to integrate with an offshore power system. A traditional approach involves expanding the onshore grid and connecting the offshore wind power with a point-to-point cable—either AC or DC, depending on distance and capacity. Grid expansion often provides local benefits, such as more available grid capacity for reginal actors. However, building new infrastructure onshore can be challenging due to public opposition. Alternatively, the offshore wind farm could be connected through a multi-terminal HVDC (MT-HVDC) system serving as both the integration of the wind power but also as the increase of the transfer capacity of the system as a whole by connecting to stronger points in the grid. Such a solution is typically more capital intensive but enables direct control of the power flow for moving energy and rescheduling power post contingency, and in weak grids the AC/DC converters can be used to stabilize the local voltage. Each solution has distinct benefits and drawbacks, and the optimal choice should be evaluated on a case-by-case basis.

The paper will first present an open, aggregated grid model of the transmission system in the Nordic synchronous area. This model can capture grid bottlenecks due to varying seasonal load and generation, using a combination of data sources in a traceable and reproduceable manner. It is validated through an hour-by-hour model simulation for the year 2024, using historical data on power flows, generation mix, and hydropower reservoir utilisation.

The grid model will be used in a case study representing a future scenario with large-scale offshore wind farms and increased power demand. The case study will assess different options for offshore wind integration and onshore/offshore grid development in Northern Norway, a region with high ambitions for offshore wind and a pressing need to strengthen the grid. A Security-Constrained Optimal Power Flow (SCOPF) model will be used that considers the optimal dispatch of resources restricted by the physical power flow during potential contingencies as well as normal operation. SCOPF models for grid models of this scale needs to be carefully engineered to ensure reasonable computational times. Previous works have shown that efficient formulations based on Power Transfer Distribution Factors (PTDFs) can achieve a good balance between computational efficiency and model accuracy for transmission systems considering a plethora of contingencies. Key economic indicators, such as grid congestion, electricity price gradients, wind power curtailment and other relevant metrics, will be computed through hour-by-hour simulations.

Using the Northern Norway-case as an illustrative example, the study will explore the potential benefits of developing an offshore grid as an alternative to traditional onshore grid expansion. The results of this study will provide quantitative insights about the potential benefits of coordinated onshore and offshore grid planning, considering power markets and security constraints.

The wind farm cable routing problem (WFCRP) falls into the combinatorial optimization (CO) category. The design variables are discrete, consisting of decisions on whether to lay an electrical power cable between each pair of nodes (turbines or substations), with the objective of minimizing the electrical network cost. A feasible network must also comply to cable's maximum current capacity, which makes the WFCRP similar to the Capacitated Minimum Spanning Tree (CMST) problem.

Recent studies have reported machine learning (ML) approaches that solve the capacitated vehicle routing problem (CVRP), which is similar to the WFCRP. It is still unclear whether ML methods can outperform specialized heuristics or meta-heuristics for the CVRP. Some works suggest that learning-enhanced heuristics can generalize to bigger instances than those used in training. They also indicate that hybridization with heuristics is instrumental in ensuring the solutions comply with all constraints.

This work introduces a learning-enhanced heuristic inspired by the Esau-Williams heuristic for the CMST. The proposed Data-Driven Heuristic (DDH) iteratively constructs solutions by selecting edges based on a data-informed appraisal. This appraisal is performed by a feed-forward neural network trained to classify whether an edge belongs to the optimal solution, using features derived from the edge and its local context during construction.

This methodology involves five steps: (1) collecting and augmenting the wind farm instances, (2) generating reference solutions, (3) training the edge appraiser, (4) integrating it into the heuristic, and (5) evaluat algorithm, and assessing the performance of the data-driven heuristic (DDH) in comparison to the reference network.

Each problem instance consists of the locations of a set of wind turbines and one electrical substation (the nodes) along with the polygonal area where cables are allowed to be laid. This area may also contain polygonal exclusion zones within the larger permitted zone. The dataset for training consists of 68 instances representing actual offshore wind farm sites, which is augmented with procedurally generated turbine positions using the same boundaries and substation position.

A total of 2800 instances (2500 for development, 300 for testing) are solved using mathematical optimization to obtain minimum-length cable networks. For each develpment instance, 10 partial solutions with random completeness levels are produced, from which the features of the remaining available edge choices are calculated. These features—20 in total—include 2 instance-level, 4 intrinsic edge-level, and 14 context-dependent edge-level features. This process produces ~20 million input-output pairs, split into training (85%) and validation (15%) sets. This data is used in a hyperparameter-optimizing framework to obtain a binary classifier called the edge appraiser.

The best-performing neural network has ~9000 trainable parameters, a binary cross entropy loss of 0.138 and a ROC AUC of 0.96. When integrated into the DDH and tested, the median excess cable length over the optimal solution is 5.3%, compared to 5.1% for the heuristic alone.

These results highlight the challenge of outperforming specialized heuristics with ML models in combinatorial optimization. However, there is still room for exploring solution-improving learning-enhanced heuristics for the WFCRP.

Offshore wind farms are a strategic asset in the energy transition, granting significant access to clean energy. Their economic viability depends on maximizing annual energy production while minimizing lifetime costs. Floating wind turbines, in particular, pose unprecedented challenges related to their dynamics, mooring system, and dynamic cables. Although some experience from deep water oil and gas operations is transferable, the placement and layout of turbines is intrinsically related to offshore wind farms and have a direct effect on the generated energy and costs. Layout optimization of offshore wind farms is fundamental to reduce the associated costs and it can help realize the full potential of offshore wind, contributing to the ambitious zero emissions targets set by different nations. This study proposes a framework for the co-optimization of wind farm layout, also called the micro-siting problem, and of the inter-array cables configuration, also called the cable routing problem. The strategy is based upon existing tools from TOPFARM, an open-source Python package. In particular, we focus on the harmonization of multiple problems and phenomena within the optimization, aiming for a more reliable estimation of the annual energy production and levelized cost of energy. We improve calculations through several additions, including an updated, floating wind turbine-specific economic model, location-specific mooring system design, turbine downtime considerations, and static pitch angle effects. Extensive research has been dedicated to several aspects of wind farm design and optimization, but combining contributions into a single framework is rare. The proposed framework increases confidence in the results by leveraging the advances made in recent years. This study is the first step towards the generalization of an accurate, powerful, and open-source tool suitable for wind farm layout design and simulation for supporting researchers and decision makers.