Ship Performance Model (NN-SPM)

NN-SPM estimates ship fuel consumption taking into consideration weather and operational conditions.

Training Data

This repository contains 'example_dataset' that was built using:

1. Weather forecast data collected from NOAA for the experimental area in the North Pacific
2. Forecasts of FuelMassFlow generated from systems that are modelling ship energy system
   • Vessel is assumed to sail at constant speed, water depth, draft
   • For each weather combination, three different COGs are contained with the resulting FuelMassFlow estimate

Input features (n=9):

• wind speed (m/s)
• wind direction (degree)
• significant wave height (m)
• peak wave period (s)
• wave direction (degree)
• primary swell height (m)
• primary swell period (s)
• primary swell direction (degree)
• course over ground (degree)

Output (n=1): Fuel Mass Flow (kg/s)

Model Architecture and Training Hyperparameters Values

A fully connected feedforward neural network.

• Layers: network has ten layers each with 10 neurons
• Activation function: Leaky ReLU
• Loss function : Mean Squared Error
• Optimizer for loss function: ADAM
• Learning rate: 0.005
• Batch size: 64
• No. of Epochs: 30