표제지
목차
Abstract 10
Ⅰ. 서론 11
1.1. 연구 배경 11
1.2. 연구 목표 및 내용 13
Ⅱ. 관련 규제 및 선행 연구 15
2.1. 국제해사기구(IMO)의 환경 규제 동향 15
2.1.1. 운항선효율지수(Energy Efficiency Existing Ship Index; EEXI) 17
2.1.2. 탄소집약도지수(Carbon Intensity Indicator; CII) 19
2.2. 선행 연구 고찰 22
2.2.1. 머신러닝 기반의 선박 연료소비량 예측 관련 연구 22
2.2.2. XAI기반의 선박 연료소비량 분석 관련 연구 24
Ⅲ. 선박 연료 소모량 예측 29
3.1. 연구 방법 소개 29
3.2. 데이터 설명 31
3.3. 데이터 정제 및 전처리 34
3.3.1. 데이터 정제 34
3.3.2. 데이터 전처리 34
3.4. XAI 관련 이론 고찰 42
3.5. XGboost 적용 선박 연료소모량 예측 모델 45
3.6. XGboost 적용 선박 연료소모량 예측 모델 평가 45
3.7. SHapley Additive exPlanation(SHAP)을 통한 모델 해석 49
Ⅳ. 결론 63
참고문헌 65
Table 1.1. 해운분야의 CO₂ 배출량 15
Table 1.2. 2019년 기준값 대비 CII 감축계수 21
Table 3.1. Vessel Specification 31
Table 3.2. Original Feature 32
Table 3.3. conversion of angles 37
Table 3.4. Transformed Feature range 37
Table 3.5. final selected variables 38
Table 3.6. Feature correlation result 42
Table 3.7. The performance evaluation of the model using the training dataset 47
Table 3.8. The performance evaluation of the model using the test datase 47
Table 3.9. Summarizing the relationship between the prediction model and features through the SHAP value plot. 51
Table 3.10. Comparison of shap value 58
Figure 1.1. 국제해사기구(IMO) 선박 온실가스배출 감축전략 17
Figure 2.1. CII 등급 및 범위 산정 방법 21
Figure 2.2. Directional Feature Transformation 25
Figure 2.3. SHAP Summary Plot 27
Figure 2.4. SHAP Feature Importance Ranking Heatmap 28
Figure 3.1. Flow chart of the study 30
Figure 3.2. Vessel Specification 31
Figure 3.3. Course of a vessel 32
Figure 3.4. LEEWAY and TIDEWAY 35
Figure 3.5. SWELL_WAVE_ANGLE 36
Figure 3.6. Vessel Draft 37
Figure 3.7. Feature distribution 40
Figure 3.8. Feature correlation matrix 41
Figure 3.9. The performance metrics of the model evaluated on the training dataset 48
Figure 3.10. The performance metrics of the model evaluated on the test dataset 48
Figure 3.11. mean absolute SHAP value 49
Figure 3.12. SHAP values(impact on model output) 50
Figure 3.13. Dependence plot of SPEED_VG and REL_WIND_SPEED 52
Figure 3.14. Dependence plot of SPEED_VG and SWELL_WAVE_PERIOD 52
Figure 3.15. Dependence plot of SPEED_VG and SWELL_WAVE_HEIGHT 52
Figure 3.16. Dependence plot of REL_WIND_SPEED and AvgDraft 52
Figure 3.17. Dependence plot of REL_WIND_SPEED and AvgDraft 52
Figure 3.18. Dependence plot of AvgDraft and Trim 52
Figure 3.19. Dependence plot of AvgDraft and WIND_WAVE_PERIOD 53
Figure 3.20. Dependence plot of AvgDraft and SPEED_VG 53
Figure 3.21. Dependence plot of SPEED_VG 53
Figure 3.22. PDP plot of SPEED_VG 53
Figure 3.23. Dependence plot of REL_WIND_SPEED 54
Figure 3.24. PDP plot of REL_WIND_SPEED 54
Figure 3.25. Dependence plot of AvgDraft 54
Figure 3.26. PDP plot of AvgDraft 54
Figure 3.27. ME1_FOC Distribution 55
Figure 3.28. ME1_FOC Plot 56
Figure 3.29. Map showing observations with high foc (red dots) 56
Figure 3.30. Force Plot with ME1_FOC greater than 2750 57
Figure 3.31. Mean of SHAP Value comparison between All and South africa 59
Figure 3.32. SHAP Force Plot 59
Figure 3.33. SHAP Waterfall Plot 60
Figure 3.34. Plots each feature in South Africa section (1) 61
Figure 3.35. Plots each feature in South Africa section (2) 62