Slide 1: Release of Wave Predictor Video and Webinar
ASX ANNOUNCEMENT 24 September 2020 Release of Wave Predictor Video and Webinar • New video explaining and promoting Carnegie’s Wave Predictor released • Blue Economy CRC webinar conducted including Wave Predictor Wave Predictor Video Carnegie is pleased to announce the release of a new video related
Slide 2: Blue Economy CRC Webinar: Machine Learning – adding power to the toolbox
Blue Economy CRC Webinar “Machine Learning – adding power to the toolbox” Yesterday Carnegie’s Data Analyst made a presentation about the Wave Predictor during a webinar coordinated by the Blue Economy Cooperative Research Centre (CRC). The topic of the webinar centred on the application of machine
Slide 3: CETO Wave Predictor
CETO Wave Predictor 23rd Sept 2020
Slide 4: Who is Carnegie Clean Energy?
Who is Carnegie Clean Energy? • Australian Stock Exchange (ASX) listed • 11,000 shareholders, mostly retail • One of the leaders in wave energy technology. 15 years of CETO development including tank testing, rapid small scale prototyping and large commercial scale prototypes and array: • Multiple s
Slide 5: What is CETO?
What is CETO? • The technology has been in development for over 15 years. • Now mainly captures the ‘heave’ and ‘surge’ motions form the waves • Sits ~2 m below the sea surface. • Installed in depths > 20 m. Commercial in Confidence and Copyright Carnegie Clean Energy Limited 2019
Slide 6: Why a wave predictor?
Why a wave predictor? • Carnegie is currently developing an advanced controller for CETO • Reinforcement Learning and Model Predictive Control are being investigated • Most advanced controllers require future knowledge of the waves • A wave predictor is required • Machine Learning allows to solve co
Slide 7: Wave Predictor development steps
Wave Predictor development steps • Development steps: • Long / Short crested linear waves: data generated on a laptop • Short crested non-linear waves: data generated on Pawsey Supercomputers • Tank testing: data generated over 10 days of testing • Real ocean Commercial in Confidence and Copyright C
Slide 8: Wave Predictor description
Wave Predictor description • Objective: Predict waves 30 s into the future, in all conditions at the deployment site • Prediction based on up-wave measurements • Dataset contains ML model input data, and output data (labels, or truth) Data Generation Training phase Optimiser WP target output Numeric
Slide 9: Feature Engineering
Feature Engineering • WP inputs: surface elevation @ wave sensor, from X s into the past until now • WP output: surface elevation @ CETO system, 30 s into the future • Distance and lookback parameters are dictated by wave travel velocity Commercial in Confidence and Copyright Carnegie Clean Energy L
Slide 10: Feature Engineering
Feature Engineering • Design parameters • Number and spacing of sensors • Distance from prediction point • Network architecture Commercial in Confidence and Copyright Carnegie Clean Energy Limited 2019
Slide 11: Wave predictor performance
Wave predictor performance • The WP performs very well on synthetic data • The WP is able to generalise to unseen sea states, provided they fall within the envelope of the training set • Currently in the process of validating it on tank data, with promising early indicators testing set Surface eleva
Slide 12: Applications outside of wave energy
Applications outside of wave energy • The WP can increase safety and operational envelope of activities where large waves are a limiting factor • Boarding / outboarding of sea vessels • Offshore lifting • The use of X-Band radar will increase the application range of the wave predictor Commercial in
Slide 13: Thank you! Questions?
Thank you! Questions ? Commercial in Confidence and Copyright Carnegie Clean Energy Limited 2019 11