Carnegie kicks off ‘Wave Predictor’ validation

Carnegie Wave Predictor

Carnegie has commenced its planned wave tank testing campaign at the Cantabria Coastal and Ocean Basin in Spain.

The trials will generate physical wave data aimed to validate Carnegie’s machine learning based Wave Predictor.

Carnegie previously announced the development of a machine learning based Wave Predictor capable of predicting the characteristics of waves that will reach the CETO Unit up to 30 seconds in the future.

This is the first product in Carnegie’s suite of control products using
artificial intelligence.

They seek to increase the energy captured from the waves and thereby increase the electric power yield of CETO Unit.

The Wave Predictor is significant for Carnegie as it enables the CETO technology to respond to wave conditions in a manner that optimises power production, thus improving the commercial performance of the technology.

Carnegie’s Wave Predictor could also benefit other applications in the marine industry, including aquaculture, offshore wind and shipping.

To develop the Wave Predictor, Carnegie’s data analysis team utilised the Pawsey Supercomputing Centre’s state of the art Magnus supercomputer to run simulations.

These generated over 250 GB of wave data used to train the neural network (an artificial brain) to predict waves in complex sea states, including directionally spread waves.

What does this tank testing campaign deliver?

This tank testing campaign will now deliver physical wave data measured by an array of equipment installed in the tank.

The tank testing campaign was delayed due to the COVID-19 pandemic.

Nevertheless, Carnegie used the opportunity to refine the size of the array of wave sensor used to perform the prediction.

Related Article

Posted: about 1 year ago

Carnegie delivers wave predictor; COVID-19 delays validation

  • Research & Development
Posted: about 1 year ago

The team has been able to reduce the number of sensors by half with no impact on prediction accuracy.

This optimisation will reduce the cost of the Wave Predictor product and make its deployment simpler, Carnegie said.

At the end of last week, the facility prepared for the testing with wave sensors installed in the basin.

Due to current travel restrictions, Carnegie cannot attend the trials, but a live stream has been set up.

This allows Carnegie to engage in the progression of the test and provide direct feedback to the operations at Cantabria.

Also, the time difference allows Carnegie to analyse the data and suggest any adjustments before the next testing starts.

The testing will span over the week, running two shift of seven hours per day.

It will see more than 200 testing runs which will produce the equivalent of 230 hours of full-scale wave data.

These test runs will cover wave period ranging from 6 to 16 seconds and wave height ranging from 1 to 8 metres.

UWA partnership

The campaign is as collaboration with the Wave Energy Research Centre from the University of Western Australia (UWA).

UWA has been working on wave prediction methods using physics-based models, in parallel with the complementary machine learning based model developed at Carnegie.

The data produced during the test will also help UWA validate their physics-based predictions algorithms.