앙상블 GRU 모델과 XAI를 이용한 하수찌꺼기 발생량 예측 및 2050 바이오가스화 정책 시나리오 분석

초록

Domestic sewage sludge management is undergoing a paradigm shift due to the saturation of sewage treatment capacity and the enforcement of resource recovery policies. This study aims to address statistical discontinuities in national data by constructing a 3-Model 2-Stage Ensemble GRU model that integrates three learning periods (Long, Mid, Short). To ensure model interpretability, SHAP (Shapley Additive Explanations) was applied to identify key influencing factors. Based on the Act on the Promotion of Production and Utilization of Biogas and national policy roadmaps, future prediction scenarios were established through 2050. SHAP analysis revealed that the determinants of sludge generation have shifted from past policy regulations to current physical infrastructure capacity. Treatment methods such as fuelization and landfilling were distinctly associated with infrastructure supply and regulatory costs, respectively. Scenario analysis showed that under the Policy_Strong scenario, which assumes an active increase in the anaerobic digestion input rate to 80% by 2050, the total sludge generated is projected to decrease significantly to 4.18 million tons, compared with 4.53 million tons in the Base scenario. Furthermore, while the Combined_Max scenario (which combines strong policy with aggressive infrastructure investment) achieved the highest fuelization rate, it also entailed a trade-off, with an increase in low-grade recycling products resulting from the quantitative expansion of infrastructure. Consequently, this study proposes a strategy of “Structural Reduction and Qualitative Upgrading,” combining source reduction with high-quality energy recovery as the optimal pathway for sustainable sludge management. © 2026, Korea Society of Waste Management. All rights reserved.

키워드