Cruise ship Energy Efficiency

Transcription

1 Cruise ship Energy Efficiency Fredrik Ahlgren Linnaeus University, Kalmar Maritime Academy Sweden MOSES 2017, EPFL

2 Agenda - What has been done. - What is to be done.

3 Fredrik Ahlgren, PhD student - Marine Engineer, Navy Officer - 75 % PhD done - PhD focusing on Energy Efficiency in Shipping

4 What has been done Waste Heat Recovery Simulations Organic Rankine Cycle F. Ahlgren, M. E. Mondejar, M. Genrup, and M. Thern, Waste Heat Recovery in a Cruise Vessel in the Baltic Sea by Using an Organic Rankine Cycle: A Case Study, J. Eng. Gas Turbines Power, vol. 138, no. 1, p , M. E. Mondejar, F. Ahlgren, M. Thern, and M. Genrup, Study of the On-route Operation of a Waste Heat Recovery System in a Passenger Vessel, Energy Procedia, vol. 75, pp , M. E. Mondejar, F. Ahlgren, M. Thern, and M. Genrup, Quasisteady state simulation of an organic Rankine cycle for waste heat recovery in a passenger vessel, Appl. Energy, pp. 1 12, Jan F. Ahlgren, M. Thern, M. Genrup and M. E. Mondejar, Energy integration of Organic Rankine Cycle Exhaust Gas recirculation and Scrubber, MARENER 2017 Energy and Exergy Analysis F. Baldi, F. Ahlgren, T.-V. Nguyen, C. Gabrielii, and K. Andersson, Energy and exergy analysis of a cruise ship, in ECOS 2015, Baldi, F., Nguyen, T., Ahlgren, F. (2016). The application of process integration to the optimisation of cruise ship energy systems : a case study. ECOS 2016 : 29th International Conference on Efficiency, Cost, Optimization, Simulation and Envirionmental Impact of Energy Systems. June F. Baldi, F. Ahlgren, F. Melino, C. Gabrielii, and K. Andersson, Optimal load allocation of complex ship power plants, Energy Convers. Manag., vol. 124, pp , 2016.

5 PAX Vessel M/S Birka Stockholm Design speed 21 knots Passengers 1800 LOA 176 m BOC 28 m Propulsion 4 Wärtsilä 6L46, kw 2 Controllable Pitch Propellers Electrical 4 Wärtsilä 6L32, kw Leisure cruises: Tax free, pool, saunas, night clubs, spa, 8 bars, 4 restaurants

6 Route Stockholm Mariehamn [18:00 05:00] [08:00 16:00] 22 h

7 Speed intervals Speed distribution, 2-knot interval % operating time percentage / % % 24% 20% 6% 8-14 knots 78 % % Vessel speed / knots

8 Operational profile

9 PAX Vessel M/S Birka Stockholm ORC 4 x W46 4 x W32

10 Method ORC study Ø Considering waste heat from exhaust of all engines, combined Ø Lower temperature 150 C, sulphur condensation Ø Dry fluids Ø Off design optimisation

11 ORC configurations a) b) T source,in (from engines) 3 2 T source,out 1 Superheater Evaporator Preheater Turbine 3' T source,in (from engines) 3 2 Superheater Evaporator Turbine T source,out Preheater 1' 4 3' b Pump 0 Intermediate loop a Condenser c T sink,in T sink,out d 4 b Pump 0 Intermediate loop a Regenerator 4' Condenser c T sink,in d T sink,out Schemata of the two ORC configurations studied in this work: a) simple ORC with intermediate condenser loop, and b) regenerated ORC with intermediate condenser loop.

12 ORC study 2 - Design Study the integration of an Organic Rankine Cycle for waste heat recovery with a Scrubber and EGR Two-stroke MAN engine with EGR for NO x reduction Wet scrubber for SO x reduction Data Two-stroke MAN 6G50ME-C95, MCR 6120 kw x 79 rpm (Courtesy of MAN Diesel & Turbo) Software used; IPSEpro - a software system for calculating heat balances and simulating processes Refprop - Reference Fluid Thermodynamic and Transport Properties Database MATLAB

13 System design EGR Exhaust receiver EGR Pre-scrubber EGR Cooler EGR Scrubber Water Mist Catcher CAC Cooler Water Mist Catcher Engine Scavenge air receiver

14 Heat sources temperature-load diagram Charge Air Cooler large differences Temperatures MAN 6G50ME-C95 CAC Temp Exhaust Exhaust EGR FW Jacket Cooling 400 Temperature Load

15 Model design cases Exhaust EGR Exhaust after turbo Air CAC FW Jacket CW

16 Net power with scrubber integration Power kw Scrubber * ORC net power -70 * Reference value from (ABS. (2013). Exhaust Gas Scrubber Systems Advisory)

17 What is to be done Waste Heat Recovery Simulations Organic Rankine Cycle Energy and Exergy Analysis Machine Learning for Energy Efficiency optimization

18 computer programming is about automation, and machine learning is all about automating automation, then automated machine learning is the automation of automating automation. - Sebastian Raschka

19 Working with... Idea from working with the data in previous studies, challenge in creating decent predictions from logged data. Not a computer scientist, Machine Learning is not easy. An Engineers approach use the best tool for the job. Auto Machine Learning Tree-based Pipeline Optimization (TPOT) R. S. Olson, R. J. Urbanowicz, J. H. Moore, and N. Bartley, Evaluation of a Tree-based Pipeline Optimization Tool for Automating Data Science, in Proceedings of the Genetic and Evolutionary Computation Conference 2016, 2016, pp Auto Sklearn M. Feurer, A. Klein, K. Eggensperger, J. Springenberg, M. Blum, and F. Hutter, Efficient and Robust Automated Machine Learning, Adv. Neural Inf. Process. Syst. 28, pp , Deep Learning TensorFlow, PyTorch. Task finding patterns, learning, unsupervised.

20 TPOT Regression, 10 generations Randomized test data

21 TPOT Regression, 10 generations Randomized test data.. zoomed in...

22 TPOT, 10 generations, exported model

23 Sklearn-Auto, running time ~1hr, model specs [( , SimpleRegressionPipeline({'imputation:strategy': 'mean', 'one_hot_encoding:use_minimum_fraction': 'True', 'preprocessor: choice ': 'feature_agglomeration', 'regressor: choice ': 'extra_trees', 'rescaling: choice ': 'standardize', 'one_hot_encoding:minimum_fraction': , 'preprocessor:feature_agglomeration:affinity': 'euclidean', 'preprocessor:feature_agglomeration:linkage': 'ward', 'preprocessor:feature_agglomeration:n_clusters': 281, 'preprocessor:feature_agglomeration:pooling_func': 'mean', 'regressor:extra_trees:bootstrap': 'False', 'regressor:extra_trees:criterion': 'mse', 'regressor:extra_trees:max_depth': 'None', 'regressor:extra_trees:max_features': 1.0, 'regressor:extra_trees:min_samples_leaf': 1, 'regressor:extra_trees:min_samples_split': 2, 'regressor:extra_trees:n_estimators': 100}, dataset_properties={ 'task': 4, 'sparse': False, 'multilabel': False, 'multiclass': False, 'target_type': 'regression', 'signed': False})), ( , SimpleRegressionPipeline({'imputation:strategy': 'mean', 'one_hot_encoding:use_minimum_fraction': 'False', 'preprocessor: choice ': 'no_preprocessing', 'regressor: choice ': 'gradient_boosting', 'rescaling: choice ': 'none', 'regressor:gradient_boosting:learning_rate': , 'regressor:gradient_boosting:loss': 'ls', 'regressor:gradient_boosting:max_depth': 8, 'regressor:gradient_boosting:max_features': , 'regressor:gradient_boosting:max_leaf_nodes': 'None', 'regressor:gradient_boosting:min_samples_leaf': 3, 'regressor:gradient_boosting:min_samples_split': 7, 'regressor:gradient_boosting:min_weight_fraction_leaf': 0.0, 'regressor:gradient_boosting:n_estimators': 169, 'regressor:gradient_boosting:subsample': }, dataset_properties={ 'task': 4, 'sparse': False, 'multilabel': False, 'multiclass': False, 'target_type': 'regression', 'signed': False})), ( , SimpleRegressionPipeline({'imputation:strategy': 'most_frequent', 'one_hot_encoding:use_minimum_fraction': 'False', 'preprocessor: choice ': 'polynomial', 'regressor: choice ': 'gradient_boosting', 'rescaling: choice ': 'standardize', 'preprocessor:polynomial:degree': 2, 'preprocessor:polynomial:include_bias': 'False', 'preprocessor:polynomial:interaction_only': 'True', 'regressor:gradient_boosting:learning_rate': , 'regressor:gradient_boosting:loss': 'ls', 'regressor:gradient_boosting:max_depth': 10, 'regressor:gradient_boosting:max_features': , 'regressor:gradient_boosting:max_leaf_nodes': 'None', 'regressor:gradient_boosting:min_samples_leaf': 13, 'regressor:gradient_boosting:min_samples_split': 11, 'regressor:gradient_boosting:min_weight_fraction_leaf': 0.0, 'regressor:gradient_boosting:n_estimators': 289, 'regressor:gradient_boosting:subsample': }, dataset_properties={ 'task': 4, 'sparse': False, 'multilabel': False, 'multiclass': False, 'target_type': 'regression', 'signed': False})), ( , SimpleRegressionPipeline({'imputation:strategy': 'most_frequent', 'one_hot_encoding:use_minimum_fraction': 'True', 'preprocessor: choice ': 'polynomial', 'regressor: choice ': 'decision_tree', 'rescaling: choice ': 'minmax', 'one_hot_encoding:minimum_fraction': , 'preprocessor:polynomial:degree': 2, 'preprocessor:polynomial:include_bias': 'True', 'preprocessor:polynomial:interaction_only': 'False', 'regressor:decision_tree:criterion': 'mse', 'regressor:decision_tree:max_depth': , 'regressor:decision_tree:max_features': 1.0, 'regressor:decision_tree:max_leaf_nodes': 'None', 'regressor:decision_tree:min_samples_leaf': 11, 'regressor:decision_tree:min_samples_split': 7, 'regressor:decision_tree:min_weight_fraction_leaf': 0.0, 'regressor:decision_tree:splitter': 'best'}, dataset_properties={ 'task': 4, 'sparse': False, 'multilabel': False, 'multiclass': False, 'target_type': 'regression', 'signed': False})), ]

24 Thank you!