Hybrid short-term runoff prediction model based on optimal variational mode decomposition, improved Harris hawks algorithm and long short-term memory network

Sun, Wei; Tian, Peng; Luo, Yuanlin; Zhang, Chu; Hua, Lei; Ji, Chunlei; Ma, Huixin

doi:10.1088/2515-7620/ac5feb

Cited by 10 publications

(4 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Currently, research on streamflow modeling is conducted by numerous independent groups, each primarily focused on developing and applying their own models and datasets. For various reasons, such as project requirements or limitations in computing resources, many researchers opt to utilize private, often small-scale streamflow datasets rather than accessing large publicly available ones, as seen in the works of Lee and Choi [48], Sun et al [49], and Li et al [50]. Additionally, given the varied technical backgrounds and time constraints of different researchers, it is impractical for them to explore all existing modeling methods with their individual datasets.…”

Section: Comparisons With a General Benchmark Methodsmentioning

confidence: 99%

Assessing Objective Functions in Streamflow Prediction Model Training Based on the Naïve Method

Lin,

Wang,

Jiang

et al. 2024

Water

View full text Add to dashboard Cite

Reliable streamflow forecasting is a determining factor for water resource planning and flood control. To better understand the strengths and weaknesses of newly proposed methods in streamflow forecasting and facilitate comparisons of different research results, we test a simple, universal, and efficient benchmark method, namely, the naïve method, for short-term streamflow prediction. Using the naïve method, we assess the streamflow forecasting performance of the long short-term memory models trained with different objective functions, including mean squared error (MSE), root mean squared error (RMSE), Nash–Sutcliffe efficiency (NSE), Kling–Gupta efficiency (KGE), and mean absolute error (MAE). The experiments over 273 watersheds show that the naïve method attains good forecasting performance (NSE > 0.5) in 88%, 65%, and 52% of watersheds at lead times of 1 day, 2 days, and 3 days, respectively. Through benchmarking by the naïve method, we find that the LSTM models trained with squared-error-based objective functions, i.e., MSE, RMSE, NSE, and KGE, perform poorly in low flow forecasting. This is because they are more influenced by training samples with high flows than by those with low flows during the model training process. For comprehensive short-term streamflow modeling without special demand orientation, we recommend the application of MAE instead of a squared-error-based metric as the objective function. In addition, it is also feasible to perform logarithmic transformation on the streamflow data. This work underscores the critical importance of appropriately selecting the objective functions for model training/calibration, shedding light on how to effectively evaluate the performance of streamflow forecast models.

show abstract

Section: Comparisons With a General Benchmark Methodsmentioning

confidence: 99%

Assessing Objective Functions in Streamflow Prediction Model Training Based on the Naïve Method

Lin,

Wang,

Jiang

et al. 2024

Water

View full text Add to dashboard Cite

show abstract

“…[42]- [44] are Lora based smart IoT systems and give performance disparities and deployment obstacles associated withNB-IoT and LoRaWANtechnologies.In [45], a hybrid short-term runoff prediction model that combines the Long Short-Term Memory(LSTM) network, the Improved Harris Hawks Optimization algorithm (IHHO), and the OptimalVariational Mode Decomposition (OVMD). The papers [46]- [47] are studies on physical and MAC layer protocols built on LoRa. [48] is a study on forest fires and different factors effecting it.…”

Section: Block Diagram and Proposed Architecturementioning

confidence: 99%

An Intelligent IoT Based Landfill Fire Prediction And Prevention System

SAKYA,

Yadav,

Shukla

et al. 2023

Preprint

View full text Add to dashboard Cite

This work focuses on India’s tallest heap of garbage at Ghazipur landfill site in New Delhi. About 2000 tonnes of garbage is being dumped at Ghazipur every day and it causes poisonous gases. Fire, sparked by methane gas coming from the garbage and takes days to extinguish. Methane belching from garbage can cause even more toxic gases. The dumping of garbage needs to be stopped for the betterment of people and the environment. The research helps detect the gases released from garbage and the importance of the project increasing due to the growth of solid waste. The detection of the risk of fire and the communication of this situation to the people could reduce the number of fires. The system also helps with automatic detection of fire in garbage-dumped areas and alerts to the authorities. This research work presents a low cost network based on Long Range(LoRa) network which is able to assess automatically the level of risk of catching fire in the solid waste. The system consists of several LoRa nodes with sensors to scale temperature, relative humidity and increase in methane level. The data coming from LoRa nodes is stored and processed in the server that is The Things Network and sends data finally to a website or an application for the analysis of the collected data using machine learning . This helps in reducing environmental pollution which results in health hazards for the people residing in the vicinity.The Simulation of the Lora Network is done in NS-3 first ,for finding the throughput and latency with different scenarios of lora nodes and gateways. The deployin stretegies are fixed after ns-3 simulation. The ns-3 simulation is done in ns-3.36.1 using lorawan module.Thereafter the prototype model is prepared using LoRA hardware and different sensors. The real time data is collected on the server for data analysis. The work have used machine learning algorithms to analyse real-time sensor data received on a network server through the LoRa Gateway to detect abnormal patterns indicative of a potential fire outbreak and to justify the ―intelligence‖ of our system. The wireless LoRa communication eliminates extensive wiring, while intelligent data analysis enhances accuracy and minimizes false alarms. Experimental evaluations demonstrate the system's effectiveness in timely fire detection and response, minimizing environmental impact and ensuring community safety. The proposed system presents a promising solution for efficient landfill fire early detection, prevention, and mitigation.

show abstract

“…Bas et al (2021) introduced the Bootstrapped Holt Method with Autoregressive Coefficients based on the Harmony Search Algorithm. Sun et al (2022) developed a hybrid short-term runoff prediction model utilizing optimal VMD, an enhanced Harris Hawks algorithm, and a Long Short-Term Memory network for forecasting runoff in four distinct study areas: Shigu, Panzhihua, Pingshan, and Zhongjiang. Mauricio, & Ostia (2023) optimized forecasting accuracy by fine-tuning the smoothing coefficients, level, trend, and seasonal parameters of the HW method using Cuckoo Search Algorithm.…”

Section: Introductionmentioning

confidence: 99%

Decomposition and Holt-Winters Enhanced by the Whale Optimization Algorithm for Forecasting the Amount of Water Inflow into the Large Dam Reservoirs in Southern Thailand

Minsan,

Minsan

2024

JCST

View full text Add to dashboard Cite

This study introduces hybrid forecasting models integrating the Whale Optimization Algorithm (WOA) with Holt-Winters (HW) and decomposition methods, applied in both additive and multiplicative models, for time series forecasting. Focusing on monthly water inflow into four dam reservoirs in Southern Thailand, the study compares these hybrid models against classical statistical models, Grid Search for Holt-Winters (Grid-HW) and Classical Decomposition (Classic-D). The analysis comprises two phases: the training dataset phase and the testing dataset phase. In the training phase, WOA demonstrates superior parameter optimization, enhancing both HW and decomposition methods, resulting in lower Mean Absolute Error (MAE) values compared to classical models. In the testing phase, performance metrics such as Root Mean Square Error (RMSE), MAE, and Symmetric Mean Absolute Percentage Error (sMAPE) are employed. The findings reveal that the Whale Optimization Algorithm with Holt-Winters (WOA-HW) and Decomposition (WOA-D) models surpass classical approaches in long-term forecasting accuracy for three dam reservoirs. Over 24 data points, the WOA with Multiplicative Holt-Winters (WOA-HWx) is optimal for Pran Buri dam, the WOA with Additive Decomposition (WOA-D+) for Bang Lang dam, and the WOA with Multiplicative Decomposition (WOA-Dx) for Kaeng Krachan dam. The Box-Jenkins approach, further refined through a Box-Cox transformation employing a natural logarithm, emerged as the superior forecasting model for Rajjaprabha dam. This model satisfied all critical statistical criteria, including normality of residuals (Anderson-Darling: 0.359, p-value: 0.433), homoscedasticity (Levene's test: 1.24, p-value: 0.274), independence (Ljung-Box test: 14.10, p-value: 0.169), and zero mean (t-test: -0.39, p-value: 0.702), establishing its robustness and reliability for forecast analysis.

show abstract

Hybrid short-term runoff prediction model based on optimal variational mode decomposition, improved Harris hawks algorithm and long short-term memory network

Cited by 10 publications

References 41 publications

Assessing Objective Functions in Streamflow Prediction Model Training Based on the Naïve Method

Assessing Objective Functions in Streamflow Prediction Model Training Based on the Naïve Method

An Intelligent IoT Based Landfill Fire Prediction And Prevention System

Decomposition and Holt-Winters Enhanced by the Whale Optimization Algorithm for Forecasting the Amount of Water Inflow into the Large Dam Reservoirs in Southern Thailand

Contact Info

Product

Resources

About