Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

From high-frequency sensors to noon reports: Using transfer learning for shaft power prediction in maritime

Akriti Sharma, Dogan Altan, Dusica Marijan, Arnbjørn Maressa

TL;DR

This paper tackles shaft power prediction for ships when only daily noon reports are available by leveraging high-frequency sensor data through cross-frequency transfer learning. A neural network baseline is trained on sensor data and then fine-tuned on noon reports with early layers frozen, yielding substantial improvements in $MAPE$, $NMAE$, and $R^2$ across sister, similar, and different vessels, and bridging the gap to sensor-based predictions. The approach demonstrates up to a 10.6% reduction in $MAPE$ for sister vessels and provides more accurate power-trend forecasts for 2024–2025, enabling practical fleet-wide energy planning without extensive sensor deployment. Overall, the method offers a scalable pathway to improve maritime energy efficiency by extracting knowledge from high-frequency data to inform low-frequency daily reports, with potential applicability to other ship-performance tasks.

Abstract

With the growth of global maritime transportation, energy optimization has become crucial for reducing costs and ensuring operational efficiency. Shaft power is the mechanical power transmitted from the engine to the shaft and directly impacts fuel consumption, making its accurate prediction a paramount step in optimizing vessel performance. Power consumption is highly correlated with ship parameters such as speed and shaft rotation per minute, as well as weather and sea conditions. Frequent access to this operational data can improve prediction accuracy. However, obtaining high-quality sensor data is often infeasible and costly, making alternative sources such as noon reports a viable option. In this paper, we propose a transfer learning-based approach for predicting vessels shaft power, where a model is initially trained on high-frequency data from a vessel and then fine-tuned with low-frequency daily noon reports from other vessels. We tested our approach on sister vessels (identical dimensions and configurations), a similar vessel (slightly larger with a different engine), and a different vessel (distinct dimensions and configurations). The experiments showed that the mean absolute percentage error decreased by 10.6 percent for sister vessels, 3.6 percent for a similar vessel, and 5.3 percent for a different vessel, compared to the model trained solely on noon report data.

From high-frequency sensors to noon reports: Using transfer learning for shaft power prediction in maritime

TL;DR

This paper tackles shaft power prediction for ships when only daily noon reports are available by leveraging high-frequency sensor data through cross-frequency transfer learning. A neural network baseline is trained on sensor data and then fine-tuned on noon reports with early layers frozen, yielding substantial improvements in , , and across sister, similar, and different vessels, and bridging the gap to sensor-based predictions. The approach demonstrates up to a 10.6% reduction in for sister vessels and provides more accurate power-trend forecasts for 2024–2025, enabling practical fleet-wide energy planning without extensive sensor deployment. Overall, the method offers a scalable pathway to improve maritime energy efficiency by extracting knowledge from high-frequency data to inform low-frequency daily reports, with potential applicability to other ship-performance tasks.

Abstract

With the growth of global maritime transportation, energy optimization has become crucial for reducing costs and ensuring operational efficiency. Shaft power is the mechanical power transmitted from the engine to the shaft and directly impacts fuel consumption, making its accurate prediction a paramount step in optimizing vessel performance. Power consumption is highly correlated with ship parameters such as speed and shaft rotation per minute, as well as weather and sea conditions. Frequent access to this operational data can improve prediction accuracy. However, obtaining high-quality sensor data is often infeasible and costly, making alternative sources such as noon reports a viable option. In this paper, we propose a transfer learning-based approach for predicting vessels shaft power, where a model is initially trained on high-frequency data from a vessel and then fine-tuned with low-frequency daily noon reports from other vessels. We tested our approach on sister vessels (identical dimensions and configurations), a similar vessel (slightly larger with a different engine), and a different vessel (distinct dimensions and configurations). The experiments showed that the mean absolute percentage error decreased by 10.6 percent for sister vessels, 3.6 percent for a similar vessel, and 5.3 percent for a different vessel, compared to the model trained solely on noon report data.

Paper Structure

This paper contains 22 sections, 9 equations, 6 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Data Fusion Approaches in Power Estimation
  4. Transfer Learning
  5. Data
  6. Feature Selection
  7. Methodology
  8. Transfer Learning
  9. Proposed Method
  10. Baseline Model
  11. Final Model
  12. Data Preprocessing
  13. Experimental Setup
  14. Evaluation metrics
  15. Results and Analysis
  16. ...and 7 more sections

Figures (6)

  • Figure 1: Overview of the proposed shaft power prediction method. Baseline models are trained using ship data from vessel S_V1, which includes noon reports and sensor data. The sensor data is fused with Copernicus (CMEMS) data. The baseline model trained on sensor data is fine-tuned using noon reports from sister vessel S_V2 through our transfer learning approach, resulting in the final model predicting shaft power.
  • Figure 2: Shaft power for S_V2 in the year 2024
  • Figure 3: Shaft power for S_V3 in the year 2024
  • Figure 4: Shaft power for S_V4 in the year 2024
  • Figure 6: Shaft power for SM_V1 in the year 2024
  • ...and 1 more figures