Optimizing Installation Cycles for Offshore Wind Farms

On- and offshore wind energy are well-proven technologies for generating sustainable, green energy. Over the last decade, the amount of energy produced by these technologies has increased exponentially.

Offshore wind farms provide a larger amount of energy than onshore farms due to their high wind exposure out in the open sea. This advantage for offshore farms, however, leads to installation challenges due to more complex construction works and hard-to-reach offshore construction sites.

Also, limited access to the real-world data makes long-term planning and forecasts difficult to implement for operative decision support.

This article proposes a cascading online-simulation framework that applies nested simulation runs, combining current forecasts and historical data, to achieve a globalized optimization. The framework uses a primary online simulation, which ties into the real-world process, to collect and process current weather measurements and forecasts.

Cascading Simulation Framework

The proposed framework uses AnyLogic simulation software and consists of two simulation models and an external manager, which manages the spawning of child simulation runs and evaluates their results.

The first (online) simulation model represents the real-world system. It collects and uses real-world weather measurements to ensure a realistic process simulation. Whenever the simulated vessel agents need to decide on a new installation cycle, this model forwards the decision request to its manager.

The manager then spawns a number of (nested) child simulation runs. Therefore, the manager instantiates each child simulation with the current state of its simulation and passes one alternative decision for the next cycle to the child. It repeats this process for each possible decision candidate. Then, the manager collects the child simulation run results and returns the best candidate to the requesting model.

The framework repeats this process for each decision point to form the first cascade of child simulation runs. Then, the framework’s design allows child simulation runs to request decisions on their own, forming additional cascades up to a predefined depth. Once the framework reaches this depth, the simulation model’s default decision strategy applies as a fallback rule to avoid an uncontrolled exponential growth. Figure 2 schematically depicts the proposed concept.

Scheme of the cascading simulation concept

Scheme of the cascading simulation concept

The framework implemented for this article uses the same AnyLogic simulation model for the online and the child simulation runs but provides different weather data for both.

The framework has been implemented in Java to connect to the AnyLogic simulation models directly. The current implementation uses Java-Reflections to modify the simulation. The use of Java-Reflections gives high versatility, allowing the manager to handle various parameters and even different simulation models with only minor changes to the code.

This article conducted an additional experiment to investigate the influence of the selected cost function on the results.

Verwandte Posts