Due to the characteristics of strong suddenness, high harmfulness, and frequent occurrence of mountain flood disasters in small watersheds, the accuracy and reliability of mountain flood forecasting are insufficient in small watersheds. This paper studies key theories and technologies, that is the uncertainty forecasting model based on hydrologic physical mechanism. We design the Bayesian Deep Learning (DL) forecasting models, it is suitable for the transfer of spatiotemporal factors caused by mountain floods to disaster probability. The models include Bayesian Linear and Bayesian Long Short-Term Memory (LSTM) model, we hope to achieve an acceptable balance between reliability (uncertainty confidence coverage) and accuracy (confidence interval width). Meanwhile, we extract effective information from multi-source and multi-dimensional hazard factors’ big data. The experiment shows the differences between Bayesian DL models, the models have long-term probability forecasting ability at both, but Bayesian LSTM is superior to Bayesian Linear in terms of reliability, accuracy and computational consumption.
Due to the many disaster-causing factors of mountain torrents, rapid changes in time and space, and rapid disasters, the traditional hydrological models and general data-driven mountainous flood forecasting models in small watershed areas have low accuracy and timeliness. Therefore, this paper proposed the time series Recurrent Neural Networks (RNN) and data processing for small mountain flood level forecasting, it includes base RNN, Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) model. We use the Tiaoxi catchment’s mountain flood data including hydrology, meteorology, geography and human activities at Zhejiang province of China. We construct multi-dimensional time series data processing and the model test system, and develop multi-dimensional disaster factor data on an hourly scale to suit the training and prediction of Deep Learning (DL)’s structure. The experiment shows that time series base RNN, LSTM and GRU models can be considered as the efficient hydrology modeling approach. The LSTM model demonstrated the best accuracy and real-time in the study area.
16 May 2024 in International Journal of Distributed Sensor Networks
Massive smart water conservancy object (WCO) need to be connected for real-time monitoring and control, which produces massive data. Unfortunately, heterogeneous data structures and semantics lead to low interoperability between WCO and management systems. To address this challenge, we propose a novel interoperability structure for a smart water conservancy system based on the Internet of Things (IoT), and the key design includes a smart WCO terminal, interoperability network, special interoperability protocol, WCO information model, and cloud platform. Universal terminal and network are the base of interoperability hardware, and special interoperability protocol and information model for interconnection of WCO are designed for smart water conservancy management system. WCO can be connected to a water conservancy Big Data processing cloud platform for interoperability applications. The application results demonstrate that our proposed WCO’s interoperability structure has obvious advantages than the general IoT at WCO interoperability. The interoperability protocol is reliable, the information model can ease interoperability and security, and the semantic dictionary is very rich and covers all semantic services of WCO.
The water levels associated with mountain floods exhibit rapid fluctuations within small watersheds, necessitating extensive data on various factors influencing such disasters to facilitate real-time forecasting. This study investigates the application of Long Short-Term Memory (LSTM) networks for mountain flood forecasting, designing a watershed-internal Knowledge Graph (KG) and Large Language Model (LLM) that encompass watershed relationships and internal information structures. We have developed a hydrological KG for the Qixi Reservoir and Qiaodongcun forecasting points located in Zhejiang Province, China, to systematically organize water conservancy data, identify significant disaster-related factors, optimize the input hydrological data, and determine the most effective combination of input data for forecasting water levels. Additionally, we have implemented Recurrent Neural Networks (RNN) and Gated Recurrent Units (GRU) for comparative analysis with LSTM. The findings indicate that the LSTM model, when integrated with the watershed-internal KG and LLM, can effectively incorporate critical elements influencing water level changes, the accuracy of the LLM-KG-LSTM model is enhanced by 3% compared to the standard LSTM model, and the LSTM series outperforms both RNN and GRU models, Our method will guide future research from the perspective of focusing on forecasting algorithms to the perspective of focusing on the relationship between multi-dimensional disaster data and algorithm parallelism.
Mountain flood water levels exhibit high variability and complexity, making them challenging to predict, and gathering long-term data of disaster-causing factors is difficult in small watersheds, the available disaster-causing variables are short-term multimodal data. In order to improve the accuracy and real-time performance of mountain flood forecasting, we import Deep Learning (DL) model for mountain flood level prediction utilizing compound time series Long Short-Term Memory (LSTM) model and multimodal short data. On the basis of LSTM model, Convolutional Neural Networks (CNN) module is added to increase the short-term window prediction ability, and Attention module is further added to improve the prediction ability of complex water level changes, forming the compound LSTM, including LSTM-CNN and LSTM-CNN-Attention model. The data of multimodal short disaster-causing factors includes hydrology, meteorology, geography, etc., by integrating the short duration time series data for compound LSTM’s input data. This study evaluates the performance of advanced models on three test data set from representative small watersheds in Zhejiang Province, China, highlighting the effectiveness of the compound LSTM model in these specific areas. The findings emphasize the benefits of utilizing the compound LSTM model for flood forecasting. In particular, the LSTM-CNN-Attention model demonstrates enhanced accuracy and real-time processing capabilities.
Flood forecasting exhibits rapid fluctuations, water level forecasting shows great uncertainty and inaccuracy in small watersheds, and the reliability and accuracy performance of traditional probability forecasting is often unbalanced. This study combined Recurrent Neural Networks (RNN) and Bayesian to establish a comprehensive forecasting model framework of RNNs-Bayesian for the forecasting of water level confidence interval, to achieve both reasonable reliability and accuracy. In the Bayesian structure, weight training was used. In the RNNs, base RNN, Long Short-term Memory (LSTM), and Gated Recurrent Unit (GRU) are used for comparative analysis, and experiments are carried out at the point of the Qixi Reservoir in a small watershed in Zhejiang Province of China. We used the multidimensional disaster data input unit for water level forecasting, including hydrology, meteorology, and geography, and 5 days of time windows for forecasting, The comprehensive reliability of LSTM-Bayesian for 0~102 hours flood reached 92.31%, and the comprehensive accuracy reached 89.15%, and confidence interval forecasting using LSTM is the best method, and achieved reasonable balance of reliability and accuracy. Overall, compound RNN could be a good alternative for forecasting hourly streamflow and extreme water level in small watersheds.