Advanced Hybrid Metaheuristic Machine Learning Models Application for Reference Crop Evapotranspiration Prediction

The building sector is the largest energy consumer accounting for 40% of global energy usage. An energy forecast model supports decision-makers to manage electric utility management. Identifying optimal values of hyperparameters of prediction models is challenging. Therefore, this study develops a novel time-series Wolf-Inspired Optimized Support Vector Regression (WIO-SVR) model to predict 48-step-ahead energy consumption in buildings. The proposed model integrates the support vector regression (SVR) and the grey wolf optimizer (GWO) in which the SVR model serves as a prediction engine while the GWO is used to optimize the hyperparameters of the SVR model. The 30-min energy data from various buildings in Vietnam were adopted to validate model performance. Buildings include one commercial building, one hospital building, three authority buildings, three university buildings, and four office buildings. The dataset is divided into the learning data and the test data. The performance of the WIO-SVR was superior to baseline models including the SVR, random forests (RF), M5P, and decision tree learner (REPTree). The WIO-SVR model obtained the highest value of correlation coefficient (R) with 0.90. The average root-mean-square error (RMSE) of the WIO-SVR was 2.02 kWh which was more accurate than those of the SVR model with 10.95 kWh, the RF model with 16.27 kWh, the M5P model with 17.73 kWh, and the REPTree model with 26.44 kWh. The proposed model improved 442.0–1207.9% of the predictive accuracy in RMSE. The reliable WIO-SVR model provides building managers with useful references in efficient energy management.

Evapotranspiration (ET) is a critical component of the water cycle, and an accurate prediction of ET is essential for water resource management, irrigation scheduling, and agricultural productivity. Traditionally, ET has been estimated using satellite-based remote sensing, which provides synoptic coverage but can be limited in spatial resolution and accuracy. Unmanned aerial vehicles (UAVs) offer improved ET prediction by providing high-resolution imagery of the Earth’s surface but are limited to a small area. Therefore, UAV and satellite images provide complementary data, but the integration of these two data for ET prediction has received limited attention. This paper presents a method that integrates UAV and satellite imagery for improved ET prediction and applies it to five crops (corn, rye grass, wheat, and alfalfa) from agricultural fields in the Walla Walla of eastern Washington State. We collected UAV and satellite data for five crops and used the combination of remote sensing models and statistical techniques to estimate ET. We show that UAV-based ET can be integrated with the Landsat-based ET with the application of integration factors. Our result shows that the Root Mean Square Error (RMSE) of daily ET for corn (Zea mays), rye grass (Lolium perenne), wheat (Triticum aestivum), peas (Pisum sativum), and alfalfa (Medicago sativa) can be improved by the application of the integration factor to the Landsat based ET in the range of (35.75–65.52%). We also explore the variability and effect of partial cloud on UAV-based ET estimation. Our findings have implications for the use of UAVs in water resource management and highlight the importance of considering multiple sources of data in ET prediction.

Predicting dam inflow is critical for human life safety, water resource management, and hydroelectric power generation. While machine learning (ML) models address complex, nonlinear hydrological problems, quantum machine learning (QML) offers greater potential to overcome classical computational limits. This study compares a hybrid quantum neural network (HQNN) with the following two classical models: bidirectional CNN-LSTM and support vector regression (SVR). These models were evaluated to predict monthly inflow to the Mile Mughan Dam, a transboundary hydroelectric and irrigation dam located on the Aras River between Azerbaijan and Iran, using a 14-year dataset (2010–2023) under two scenarios. In total, 70% of data was used for training and 30% for testing. The first scenario encompassed meteorological variables plus three months of inflow lags, and the second included inflow lags only. Model performance was assessed using Coefficient of Determination (R2), Root Mean Squared Error (RMSE), Nash–Sutcliffe efficiency (NSE), Mean Absolute Percentage Error (MAPE), and graphical plots. HQNN showed superior performance across all metrics. In Scenario 1, HQNN achieved R2 = 0.915, RMSE = 37.318 MCM, NSE = 0.908, MAPE = 8.343%; CNN-BiLSTM had R2 = 0.867, RMSE = 46.506 MCM, NSE = 0.858, MAPE = 10.795%; SVR had R2 = 0.846, RMSE = 52.372 MCM, NSE = 0.821, MAPE = 12.772%. In Scenario 2, HQNN maintained strong performance (R2 = 0.855, RMSE = 48.56 MCM, NSE = 0.845, MAPE = 9.979%) and outperformed CNN-BiLSTM (R2 = 0.810, RMSE = 56.126 MCM, NSE = 0.793, MAPE = 11.456%) and SVR (R2 = 0.801, RMSE = 60.336 MCM, NSE = 0.761, MAPE = 12.901%). In Scenario 1 and Scenario 2, HQNN increased the prediction accuracy by 19.76% and 13.47%, respectively, compared to the CNN-BiLSTM model. These results confirm HQNN’s reliability in both multivariate and univariate modeling.

Emerging Harris Hawks Optimization based load demand forecasting and optimal sizing of stand-alone hybrid renewable energy systems– A case study of Kano and Abuja, Nigeria

Short-term forecasting of streamflow by integrating machine learning methods combined with metaheuristic algorithms

Improving estimation of evapotranspiration during soil freeze-thaw cycles by incorporating a freezing stress index and a coupled heat and water transfer model into the FAO Penman-Monteith model

Evapotranspiration (ET) estimation is a primary problem for irrigation engineers and hydraulic designers because it is an important part of hydrologic cycle. Even it is non-negligible in hydraulic design calculations, it is not clear enough to estimate or calculate ET. There are some meteorological parameters which effect ET directly or indirectly such as Relative Humidity (RH), Solar Radiation (SR), Air Temperature (AT) and Wind Speed (U). In this study authors used Adaptive Neuro-Fuzzy Inference System (ANFIS) for prediction of ET and results are compared with Penman FAO 56 empirical formula. 1158 daily AT, SR, RH and U values are used to train ANFIS model and 385 daily values are used to test it. ANFIS model determination coefficient with daily observed ET values found as 0.909. Also test set values are used to calculate Penman FAO 56 formula and the determination coefficient of Penman FAO 56 with daily observed ET values found as 0.857. For the comparison of the ANFIS model and Penman FAO 56 formula results Mean Square Error (MSE) and Mean Absolute Error (MAE) are computed. According to the comparison it is understood that ANFIS model has better performance than Penman FAO 56 empirical formula for the prediction of daily ET.

Stream-flow forecasting using extreme learning machines: A case study in a semi-arid region in Iraq

This research was done to build computational models for estimating solubility of hydrogen (S) in a given system based on the inputs of temperature (T) and pressure (P). In fact, multiples models were built considering double inputs and single output. To achieve this, three different regression algorithms were trained and tested on a comprehensive dataset using the optimization technique of Grey Wolf Optimization (GWO). The three models used were Elastic Net, Support Vector Regression (SVR), and Automatic Relevance Determination (ARD) Regression. R2 score, root mean square error (RMSE), and mean absolute error were just few of the measures used to assess the models' overall efficacy (MAE). The Elastic Net model achieved the highest R2 score of 0.965 with an RMSE of 8.84×10−2 and an MAE of 7.79×10−2. The SVR technique also performed well, indicated R2 of 0.949, an RMSE of 9.81×10−2, and an MAE of 7.79×10−2. The ARD Regression model revealed an R2 of 0.905 with an RMSE of 1.53×10−1 and an MAE of 1.32×10−1. These findings highlight the potential of machine learning models and metaheuristic algorithms such as GWO in accurately estimating gas solubility in oil, which can have significant implications for various industrial applications.

The phenomenon of evapotranspiration (ET) is closely linked to the issue of water scarcity, as it involves water loss through both evaporation and plant transpiration. Accurate prediction of evapotranspiration is of utmost importance in the strategic planning of agricultural irrigation, effective management of water resources, and precise hydrological modeling. The current investigation aims to predict the monthly ET values in the Elazig province by developing an artificial neural network (ANN) model utilizing the Levenberg–Marquardt method. Consequently, the values of temperature, precipitation, relative humidity, solar hour, and mean wind speed were utilized in forecasting evapotranspiration values by implementing ANN algorithms. This research makes a valuable contribution to the existing body of literature by utilizing an ANN model developed with the Levenberg–Marquardt method to estimate evapotranspiration. It has been discovered that evapotranspiration values are impacted by various factors such as temperature (minimum, average, maximum), relative humidity (minimum, average, maximum), wind speed, solar hour, and precipitation values, which are taken into consideration for prediction. The findings indicated that Elazig, Keban, Baskil, and Agin sites had R values of 0.9995, 0.9948, 0.9898, and 0.9994 in the proposed model. It was found that Elazig’s MAPE ranged from 0 to 0.2288, Keban’s was 0.0001 to 0.3703, Baskil’s was between 0 and 0.4453, and Agin’s was both 0 and 0.2784. The findings obtained from the proposed model are compatible with evapotranspiration values computed from the Hargreaves method (R2 = 0.996). The study’s findings provide significant insights for planners and decision-makers involved in the planning and managing water resources and agricultural irrigation.

Rainfall being one of the key components of the hydrological cycle contributes significantly to assessing flood and drought events. Forecasting rainfall events is vital in field of hydrology and meteorology. A wide range of practical problems has been resolved utilizing multilayer perceptron (MLP). Optimization algorithms assist neural networks in selecting appropriate weights and obtains accurate results. In this study, grey wolves optimization (GWO) meta-heuristic algorithm is used for training MLP for improving accurateness of rainfall forecasting in one rain-gauge station (Silchar) of Cachar district, Assam, India. Performance of hybrid MLP-GWO algorithm is assessed against conventional MLP model using root mean square error (RMSE), coefficient of determination (R2) and Nash–Sutcliffe efficiency (NSE). Input parameters such as monthly average temperature, relative humidity, and rainfall data are considered for a time period of 1980–2019 for rainfall forecasting. Results showed that MLP-GWO3 with R2—0.9816, RMSE—38.54, and NSE—0.985 presented the most accurate forecasting in Silchar station. Final results specified that GWO algorithm improved the accurateness of standalone MLP model and can be suggested for forecasting monthly rainfall.KeywordsRainfallSicharMLP-GWO

Flooding is a recurrent hazard occurring worldwide, resulting in severe losses. The preparation of a flood susceptibility map is a non-structural approach to flood management before its occurrence. With recent advances in artificial intelligence, achieving a high-accuracy model for flood susceptibility mapping (FSM) is challenging. Therefore, in this study, various artificial intelligence approaches have been utilized to achieve optimal accuracy in flood susceptibility modeling to address this challenge. By incorporating the grey wolf optimizer (GWO) metaheuristic algorithm into various models—including recurrent neural networks (RNNs), support vector regression (SVR), and extreme gradient boosting (XGBoost)—the objective of this modeling is to generate flood susceptibility maps and evaluate the variation in model performance. The tropical Manimala River Basin in India, severely battered by flooding in the past, has been selected as the test site. This modeling utilized 15 conditioning factors such as aspect, enhanced built-up and bareness index (EBBI), slope, elevation, geomorphology, normalized difference water index (NDWI), plan curvature, profile curvature, soil adjusted vegetation index (SAVI), stream density, soil texture, stream power index (SPI), terrain ruggedness index (TRI), land use/land cover (LULC) and topographic wetness index (TWI). Thus, six susceptibility maps are produced by applying the RNN, SVR, XGBoost, RNN-GWO, SVR-GWO, and XGBoost-GWO models. All six models exhibited outstanding (AUC above 0.90) performance, and the performance ranks in the following order: RNN-GWO (AUC: 0.968) > XGBoost-GWO (AUC: 0.961) > SVR-GWO (AUC: 0.960) > RNN (AUC: 0.956) > XGBoost (AUC: 0.953) > SVR (AUC: 0.948). It was discovered that the hybrid GWO optimization algorithm improved the performance of three models. The RNN-GWO-based flood susceptibility map shows that 8.05% of the MRB is very susceptible to floods. The modeling found that the SPI, geomorphology, LULC, stream density, and TWI are the top five influential conditioning factors.

Evaluation and development of shale gas deposits need exact calculation of adsorbed gas concentration. However, gas adsorption and desorption experiments are time-consuming and costly. Models based on physics and empirical correlations cannot anticipate these experiments. Based on geological factors, this study intends to construct a cost-effective and accurate machine-learning model to estimate adsorbed shale gas. In this study, 601 data points from shale gas reserves were utilized. These include reservoir temperature (T,°C), total organic carbon (TOC, wt%), vitrinite reflectance (Ro,%), Langmuir pressure (PL), and volume. Based on Support Vector regression (SVR), Particle Swarm Optimization (PSO), Grey Wolf Optimization (GWO), and Sparrow Search Algorithms (SSA) were created, trained, and evaluated. The accuracy of the different models is calculated by correlation coefficient (R2), root mean square error (RMSE), average absolute relative deviation (AARD), and the duration of time each specified model (PSO-SVR, GWO-SVR, and SSA-SVR) takes to evaluate the effectiveness of each prediction model. The results showed that three optimization models (PSO-SVR, GWO-SVR, and SSA-SVR) can make good predictions. However, the PSO-SVR model is the most accurate at predicting Langmuir pressure and volume and takes the least amount of time, with RMSE and R2 values of 0.09990 and 0.9605, respectively. The GWO-SVR has an estimated RMSE of 0.1092 and R2 of 0.9529, whereas the SSA-SVR has 0.1264 and 0.9368 for Langmuir volume. PSO-SVR had the lowest time requirement for Langmuir pressure data inputs, with RMSE and R2 of 0.5017 and 0.9306, respectively. The findings indicate that all models can accurately forecast adsorbed gas, with the PSO-SVR model performing somewhat better than the GWO- and SSA-SVR models

Global solar radiation prediction by ANN integrated with European Centre for medium range weather forecast fields in solar rich cities of Queensland Australia

Testing DRAINMOD-FOREST for predicting evapotranspiration in a mid-rotation pine plantation