Mass Conservative Time-Series GAN for Synthetic Extreme Flood-Event Generation: Impact on Probabilistic Forecasting Models

Stats, Journal Year: 2024, Volume and Issue: 7(3), P. 808 - 826

Published: Aug. 3, 2024

The lack of data on flood events poses challenges in management. In this paper, we propose a novel approach to enhance flood-forecasting models by utilizing the capabilities Generative Adversarial Networks (GANs) generate synthetic events. We modified time-series GAN incorporating constraints related mass conservation, energy balance, and hydraulic principles into model through appropriate regularization terms loss function using conservative LSTM generator discriminator models. way, can improve realism physical consistency generated extreme flood-event data. These ensure that adhere fundamental hydrological characteristics, enhancing accuracy reliability risk-assessment applications. PCA t-SNE are applied provide valuable insights structure distribution data, highlighting patterns, clusters, relationships within aimed use supplement original train probabilistic neural runoff for forecasting multi-step ahead t-statistic was performed compare means TimeGAN with results showed two datasets were statistically significant at 95% level. integration GAN-generated real improved robustness autoencoder model, enabling more reliable predictions pilot study, trained augmented dataset from shows higher NSE KGE scores = 0.838 0.908, compared 0.829 0.90 sixth hour ahead, indicating 9.8% multistep-ahead alone. training improves model’s ability achieve reduced Prediction Interval Normalized Average Width (PINAW) interval forecasting, yet enhancement comes trade-off Coverage Probability (PICP).

Language: Английский

---

Mitigation imbalance distribution: Data augmentation of local small sample for building electricity load in time-series generative adversarial network

Journal of Building Engineering, Journal Year: 2024, Volume and Issue: 99, P. 111549 - 111549

Published: Dec. 9, 2024

Language: Английский

---

Enhancing Flood Resilience: Streamflow Forecasting and Inundation Modeling in Pakistan

Published: Dec. 7, 2023

Climatic changes have increased the frequency of natural disasters, and Pakistan, as a developing nation, is facing severe challenges in coping with floods, which devastatingly impacted people's livelihoods. In 2022, floods affected over 33 million people, resulting more than 1730 deaths, according to World Bank. Flood prediction critical research area can aid saving lives, crops, livestock, money. This study employs machine learning techniques provide accurate reliable flood forecasts for Pakistan. Specifically, Support Vector Machine (SVM) Artificial neural network (ANN) are utilized this prediction. Historical data encompassing rainfall, temperature, water level, topographic information, land cover Pakistan collected split into 75% model training 25% testing. Additionally, information employed create inundation maps. The findings highlight three factors that play pivotal role predicting flood-sensitive areas: slope, distance river, river. combined exhibited areas under curve values 0.94 0.95 testing phases, respectively. These results demonstrate efficacy SVM ANN integration precise forecasting contributing enhancing resilience region.

Language: Английский

---

Probabilistic multi-step ahead streamflow forecast based on deep learning

at - Automatisierungstechnik, Journal Year: 2024, Volume and Issue: 72(6), P. 518 - 527

Published: June 1, 2024

Abstract The use of deep learning methods for fluvial flood forecasting is rapidly gaining traction, offering a promising solution to the challenges associated with accurate yet time-consuming numerical models. This paper presents two physics-inspired approaches specifically designed forecasting, each embracing different principles: centralized and federated learning. model utilizes an Encoder-Decoder technique handle input data varying types scales, while based on node-link graph seq2seq internal model. Both models are enhanced probabilistic head account inherent uncertainty in streamflow forecasts. objective these address limitations traditional leveraging potential improve speed, accuracy, scalability forecasting. To validate their effectiveness, were tested across cases. findings from approach emphasize importance catchment clustering before utilization demonstrate models’ ability generalize effectively catchments similar properties. On other hand, results method highlight model’s reliance test set falling within range training (Average NSE KGE sixth hour ahead 0.88 0.78, respectively). this limitation, suggests development future, such as or using Generative Adversarial Networks, generate highly extreme events, particularly context changing climate. implemented flexible operational framework open standards, ensuring adaptability usability various settings.

Language: Английский

---

Modeling Dissolved Oxygen Under Data Scarcity Situation Using Time-Series Generative Adversarial Network Combined with Long Short-Term Memory Network

Published: Jan. 1, 2024

Language: Английский

---