Prediction and optimisation of fuel consumption for inland ships considering real-time status and environmental factors
Introduction
Ships’ fuel consumption occupies a major part of ship operating costs (Leifsson et al., 2008). It also reflects the operating status of ship engines (Schaub et al., 2019), which makes it an important monitoring variable in modern intelligent ship systems (Li et al., 2019) and a key parameter to control in unmanned ship navigation systems (Wright, 2019). On the other hand, the ship fuel consumption is also closely related to exhaust emissions and is an important indicator for pollution research and environmental monitoring (Van et al., 2019; Hansen et al., 2020). Therefore, the ship fuel consumption has become an important research topic for many scholars and practitioners.
Over the past decade, ship speed optimisation has been considered as an effective approach to improve the energy efficiency and reduce the fuel consumption. Therefore, some researchers have focused on the speed optimisation of the whole voyage for reducing the fuel consumption (Fagerholt et al., 2010, 2015; Wang and Meng, 2012; Psaraftis et al., 2014; Wen et al., 2017; Li et al., 2018; Du et al., 2019). However, the speed of a ship depends to a large extent on the speed of the ship engines, the engines’ running conditions and the environmental conditions. All of these factors have a significant effect on the ship fuel consumption and fleet cost during the voyage. Wang et al. (2016) established an approach for real-time optimisation of ship energy efficiency during the working condition in a short distance ahead of the ship and achieved real-time optimisation under different navigation conditions. Sheng et al. (2019) developed a mixed-integer convex cost-minimisation method for determination of optimal vessel speeds and fleet size. However, they all took the ship speed as the decision variable, where the ship speed is provided by the output power of engines, and is also closely related to navigational conditions. To achieve a suggested speed of a ship in the varying navigational conditions, one may need to constantly change the engine running speed, which may cause more fuel consumption than expected.
In recent years, intelligent sensing devices with high acquisition rates are more and more widely used in modern ships, and many real-time and continuous data collection systems have been developed. Using the new systems, a large number of multi-source monitoring data, including longitude, latitude, Speed Over Ground (SOG), Course Over Ground (COG), engine speed, engine temperature, voyage mileage, reserve fuel and bunker fuel, have been collected. This provides abundant fundamental data for fuel consumption prediction, energy efficiency optimisation and emission reduction (Satpathi et al., 2017; Huang et al., 2018; Wang et al., 2019).
For fuel consumption prediction, there have been some methods and models in the existing literature. Beşikçi et al. (2016) tried to reduce the ship speed and predict ship fuel consumption for various operational conditions through employment of an Artificial Neural Network (ANN), where seven input variables, including ship speed, revolutions per minute, mean draft, trim, cargo quantity on board, wind and sea effects, were used. Coraddu et al. (2017) compared three different approaches, a White Box Model, a Black Box Model and a Gray Box Model (GBM), in the prediction of the fuel consumption based on data measured by on-board automation systems. Wang et al. (2018) proposed a prediction model for the ship fuel consumption on the basis of the Least Absolute Shrinkage and Selection Operator (LASSO) regression algorithm. It used the dataset of ship reports, which includes the information about length of overall, beam, SOG, Beaufort scale and swell height. Yuan and Nian (2018) developed a Gaussian Process (GP) meta-model to predict the ship fuel consumption for different scenarios in consideration of the operational conditions’ effects, which involves speed, draft, trim, wind speed, wind direction, wave height and wave direction. Yang et al. (2019) proposed a genetic-algorithm-based GBM for the ship fuel consumption prediction using ship speed and Beaufort number. Gkerekos et al. (2019) presented a comparison of multiple data-driven regression algorithms for predicting the main engine fuel oil consumption, including Support Vector Machines, Random Forest Regressors, Extra Trees Regressors and ANNs. They considered vessel speed, engine speed and sea conditions as input variables.
However, the existing research has some limitations. (1) Limited input variables were used for fuel consumption prediction. There is a lack of some important variables, such as engine temperature, water speed and wind direction. (2) There is a lack of detailed analysis about the trajectory characteristics and geographic environment when predicting fuel consumption. (3) In the optimisation of fuel consumption and total cost, only the vessel speed was used as the decision variable, while the more directly-related and controllable variable, engine speed, was not employed. (4) The environmental factors were not taken into consideration when optimising the fuel consumption and the total running cost.
To solve the above problems, this work collects various monitoring data of ship sailing by multi-source sensors. After specific data processing and analysis, the real-time fuel consumption rate of ships is calculated and the feature variables that are most correlated with fuel consumption are obtained. The prediction model for real-time fuel consumption rate is then constructed based on the Long Short-Term Memory (LSTM) network, which is verified by the measured data and compared with some traditional regression methods, Back Propagation Neural Networks (BPNNs) and other Recurrent Neural Networks (RNNs). The prediction model of SOG will be used for fuel consumption and cost optimisation, which is also built by the LSTM network. Finally, an optimisation algorithm Reduced Space Searching Algorithm (RSSA) is used to minimise the fuel consumption and the total cost of a voyage. RSSA is a nature-inspired heuristic technique that tries to switch and zoom in/out the targeted search space to speed up the searching process and jump out from local optima. It has been verified to be able to find optimal solutions fast and accurately, and outperform some other well-known heuristic optimisation algorithms (Zhang and Mahfouf, 2010). The whole research framework is as shown in Fig. 1.
The rest of the paper is organised as follows. First, the detailed multi-source data processing and analysis are presented in Section 2, where a real-time fuel consumption calculation method is proposed and the correlation between multiple variables is analysed. Then, the prediction model of real-time fuel consumption rate is constructed in Section 3. Detailed experiments are carried out to optimise the fuel consumption and the total voyage cost of inland ships in Section 4. Finally, conclusions are drawn in Section 5. Table 1 summarises the abbreviations used in the paper.
Section snippets
Data collection
The data studied in this work came from the cargo ship sailing on the Yangtze River trunk, which was equipped with two engines rated at 735 kW. The main parameters of the ship are shown in Table 2. The raw data were collected by the multi-source sensors installed on the ship, such as Global Positioning System (GPS), Automatic Identification System (AIS), fuel sensor, speed sensor, temperature sensor and others. The collected data include IMO, ship name, date, time, longitude, latitude, SOG,
Modelling method
The LSTM (Long Short-Term Memory) network is an advance RNN model, which was proposed to solve the problem of gradient dispersion in the conventional RNN model. As shown in Fig. 6, LSTM has two transmission states and , one gate control signal and three gate control states , and , where , and represent three different control gates: input gate, forget gate and output gate. is time variable, is the input vector and is the output vector. is named “cell state”, which
Optimisation of fuel consumption and voyage cost
On the basis of the developed models, optimisation of fuel consumption and voyage cost of inland ships is then conducted, considering different environmental conditions. In this work, two optimisation problems are targeted, to minimise the fuel consumption of the whole voyage and to minimise the total cost of the whole voyage.
Conclusions
In this paper, based on the multi-source data composed of monitoring data and hydrological data, the real-time fuel consumption of inland ships has been analysed and modelled, and the optimisation of fuel consumption and the total cost for a whole voyage has been performed. The multi-source monitoring data have been processed to delete abnormal data and retain the ship's fuel information to the maximum extent. A method has been proposed to calculate the accurate real-time fuel consumption rate
CRediT authorship contribution statement
Zhi Yuan: Methodology, Data curation, Software, Writing - original draft. Jingxian Liu: Conceptualization, Methodology, Supervision. Qian Zhang: Formal analysis, Methodology, Writing - review & editing. Yi Liu: Data curation, Formal analysis. Yuan Yuan: Data curation. Zongzhi Li: Formal analysis, Visualization.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This work is supported by National Key Research and Development Program of China (Grant No. 2018YFC1407400), the National Natural Science Foundation of China (NSFC) (Grant No. 51709219, 51609195 and 51809207), the China Scholarship Council (CSC) (Grant No. 201906950086) and the Qingdao Research Institute of Wuhan University of Technology (Grant No. 2019A02).
References (38)
- et al.
Vessels fuel consumption forecast and trim optimisation: a data analytics perspective
Ocean Eng.
(2017) - et al.
Imprecise knowledge based design and development of titanium alloys for prosthetic applications
Journal of the Mechanical Behavior of Biomedical Materials
(2016) - et al.
Two-phase optimal solutions for ship speed and trim optimization over a voyage using voyage report data
Transp. Res. Part B Methodol.
(2019) - et al.
Maritime routing and speed optimization with emission control areas
Transport. Res. C Emerg. Technol.
(2015) - et al.
Machine learning models for predicting ship main engine Fuel Oil Consumption: a comparative study
Ocean Eng.
(2019) - et al.
Machine learning estimates of plug-in hybrid electric vehicle utility factors
Transport. Res. Transport Environ.
(2019) - et al.
Making shipping more carbon-friendly? Exploring ship energy efficiency management plans in legislation and practice
Energy Research & Social Science
(2020) - et al.
Integrating multi-source maritime information to estimate ship exhaust emissions under wind, wave and current conditions
Transport. Res. Transport Environ.
(2018) - et al.
A wavelet-Particle swarm optimization-Extreme learning machine hybrid modeling for significant wave height prediction
Ocean Eng.
(2020) - et al.
Grey-box modeling of an ocean vessel for operational optimization
Simulat. Model. Pract. Theor.
(2008)
Model of speed optimization of oil tanker with irregular winds and waves for given route
Ocean Eng.
An improved recurrent neural network for unmanned underwater vehicle online obstacle avoidance
Ocean Eng.
End-to-end navigation for autonomous underwater vehicle with hybrid recurrent neural networks
Ocean Eng.
Ship speed optimization: concepts, models and combined speed-routing scenarios
Transport. Res. C Emerg. Technol.
Optimal vessel speed and fleet size for industrial shipping services under the emission control area regulation
Transport. Res. C Emerg. Technol.
Genetic algorithm based design optimization of a passive anti-roll tank in a sea going vessel
Ocean Eng.
Sailing speed optimization for container ships in a liner shipping network
Transport. Res. E Logist. Transport. Rev.
Real-time optimization of ship energy efficiency based on the prediction technology of working condition
Transport. Res. Transport Environ.
Predicting ship fuel consumption based on LASSO regression
Transport. Res. Transport Environ.
Cited by (46)
Harnessing the power of Machine learning for AIS Data-Driven maritime Research: A comprehensive review
2024, Transportation Research Part E: Logistics and Transportation ReviewDevelopment of a ship performance model for power estimation of inland waterway vessels
2023, Ocean Engineering