다변량 입력이 딥러닝 기반 저수율 예측에 미치는영향 분석과 중장기 예측 방안
Analyzing the Impact of Multivariate Inputs on Deep Learning-Based Reservoir Level Prediction and Approaches for Mid to Long-Term Forecasting
0
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
지역 저수지들은 농업용수 공급의 중요한 수원공으로 가뭄과 같은 극단적 기후 조건을 대비하여 안정적인 저수율 관리가 필수적이다. 저수율예측은 국지적 강우와 같은 지역적 기후 특성뿐만 아니라 작부시기를 포함하는 계절적 요인 등에 크게 영향을 받기 때문에 적절한 예측 모델을선정하는 것만큼 입/출력 데이터 간 상관관계 파악이 무엇보다 중요하다. 이에 본 연구에서는 1991년부터 2022년까지의 전라북도 400여 개 저수지의광범위한 다변량 데이터를 활용하여 각 저수지의 복잡한 수문학·기후학적 환경요인을 포괄적으로 반영한 저수율 예측 모델을 학습 및 검증하고,각 입력 특성이 저수율 예측 성능에 미치는 영향력을 분석하고자 한다. 신경망 구조에 따른 저수율 예측 성능 개선이 아닌 다변량의 입력 데이터와예측 성능 간의 상관관계에 초점을 맞추기 위하여 실험에 사용된 예측 모델로 합성곱신경망 또는 순환신경망과 같은 복잡한 형태가 아닌 완전연결계층,배치정규화, 드롭아웃, 활성화 함수 등의 조합으로 구성된 기본적인 순방향 신경망을 채택하였다. 추가적으로 대부분의 기존 연구에서는 하루 단위의단기 예측 성능만을 제시하고 있으며 이러한 단기 예측 방식은 10일, 한 달 단위 등 중장기적 예측이 필요한 실무환경에 적합하지 않기 때문에,본 연구에서는 하루 단위 예측값을 다음 입력으로 사용하는 재귀적 방식을 통해 최대 한 달 뒤 저수율 예측 성능을 측정하였다. 실험을 통해 예측기간에 따른 성능 변화 양상을 파악하였으며, Ablation study를 바탕으로 예측 모델의 각 입력 특성이 전체 성능에 끼치는 영향을 분석하였다.
Abstract
Local reservoirs are crucial sources for agricultural water supply, necessitating stable water level management to prepare for extremeclimate conditions such as droughts. Water level prediction is significantly influenced by local climate characteristics, such as localizedrainfall, as well as seasonal factors including cropping times, making it essential to understand the correlation between input and outputdata as much as selecting an appropriate prediction model. In this study, extensive multivariate data from over 400 reservoirs inJeollabuk-do from 1991 to 2022 was utilized to train and validate a water level prediction model that comprehensively reflects the complexhydrological and climatological environmental factors of each reservoir, and to analyze the impact of each input feature on the predictionperformance of water levels. Instead of focusing on improvements in water level performance through neural network structures, thestudy adopts a basic Feedforward Neural Network composed of fully connected layers, batch normalization, dropout, and activationfunctions, focusing on the correlation between multivariate input data and prediction performance. Additionally, most existing studiesonly present short-term prediction performance on a daily basis, which is not suitable for practical environments that require mediumto long-term predictions, such as 10 days or a month. Therefore, this study measured the water level prediction performance up to onemonth ahead through a recursive method that uses daily prediction values as the next input. The experiment identified performancechanges according to the prediction period and analyzed the impact of each input feature on the overall performance based on an Ablationstudy
참고문헌 (17)
- H. Lim, "Prediction of pollution loads in agricultural reservoirs using LSTM algorithm : Case study of reservoirs in Nonsan City", Korean Journal of Agricultural Science, 49(2), pp.193~202 (2022)
- J. Jung, "A study on the use of geospatial information-based simulation for preemptive response to water disasters in agricultural land", Smart Media Journal, 11(7), pp.52~60 (2022)
- Y. Seo, "Reservoir water level forecasting using machine learning models", Journal of the Korean Society of Agricultural Engineers, 59(3), pp.97~110 (2017)
- S. Jung, "Prediction of river water level using deep-learning open library", Journal of the Korean Society of Hazard Mitigation, 18(1), pp.1~11 (2018)
- H. Sak, "Long short-term memory recurrent neural network architectures for large scale acoustic modeling" (2014)
- H. Han, "Application of sequence to sequence learning based LSTM model(LSTM"s2s)for forecasting dam inflow", Journal of Korea Water Resources Association, 54(3), pp.157~166 (2021)
- M. Yang, "Anomaly detection in reservoir water level data using the LSTM model based on deep learning", Journal of the Korean Society of Hazard Mitigation, 21(1), pp.71~81 (2021)
- Y. Seong, "Flow rate prediction at Paldang Bridge using deep learning models", Journal of Korea Water Resources Association, 55(8), pp.565~575 (2022)
- Joh. Sunguk, "Prediction of Water Storage Rate for Agricultural Reservoirs Using Univariate and Multivariate LSTM Models", Korean Journal of Remote Sensing, 39(5), pp.1125~1134 (2023)
- M. Das, "A probabilistic nonlinear model for forecasting daily water level in reservoir", Water Resources Management, 30, pp.3107~3122 (2016)
- "RAWRIS"[URL]
- "KMA Weather Data Service"[URL]
- H. L. Penman, "Natural evaporation from open water, bare soil and grass", Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences, 193(1032), pp.120~145 (1948)
- A. Paszke, "Pytorch: An imperative style, highperformance deep learning library", Advances in Neural Information Processing Systems, 32 (2019)