인공지능 기반 콜드체인 배송차량 에너지 소비량 예측

AI-Based Energy Consumption Predictions for Cold-chain Delivery Vehicles

최근 글로벌 물류 경쟁 환경의 발전에 따라 물류 고도화 시스템 개발의 필요성이 증가하고 있다. 특히 콜드체인 배송 차량에 대하여 미국, 유럽 등 주요 선진국을 중심으로 환경규제 및 에너지 절감을 요구하고 있으며, 지속가능한 물류센터 운영을 위해서는 이를 해결할 방안이 필요한 실정이다. 본 논문에서는 콜드체인 물류 환경의 경쟁력 제고와 효율적인 운영을 위해 Long Short-Term Memory (LSTM)를 통해 콜드체인 배송차량의 에너지 소비량을 예측하고, 인공지능 기술을 기반으로 스마트 물류센터 내에서 수집된 데이터를 수집·분석하여 의미 있는 정보를 추출·활용하기 위한 데이터 분석 체계 설계를 제안한다. 제안된 시스템은 콜드체인 물류 환경의 효과적 운영을 지원하며, 지속가능한 콜드체인 네트워크 구축 및 관리시스템 구축에 활용이 될 것으로 예상된다.

Purpose : This study proposes a framework for a cold-chain logistics control system, and presents a method for extracting and utilizing data collected in real-time. Besides, an AI-based data learning used to predict energy consumed by cold-chain delivery vehicles is described with a case study. Research design, data and methodology : In this paper, the energy consumption of cold-chain delivery was predicted using Long Short-Term Memory (LSTM) to enhance competitiveness and efficient operations in a cold-chain logistics environment. Data on dairy product delivery were used in a case study. In total, 539 sets of data were acquired by collecting data every five minutes from five vehicles within 2,695 minutes. Results : The LSTM model seemed to fit actual energy usage data as the epochs of the model increased. Hence, the model is expected to be more effective in predicting energy consumption at epoch settings of 1,000 or more. The validation losses for each 100, 1,000, and 10,000 epochs were 0.72802, 0.01571, and 0.00546, respectively. Conclusions : In this study, the framework of an integrated management system for smart logistics centers was proposed. In addition, a method for extracting and utilizing data collected in real-time was presented. In particular, this study contributes to maintaining temperature and enhancing energy efficiency by predicting the energy consumption of cold-chain delivery vehicles through AI-based data learning.