support vector regression

Wastewater treatment plants (WWTPs) play a vital role in protecting public health and the environment, but the treatment process itself is energy-intensive. Among the various stages, pumping stations stand out as major energy consumers, responsible for lifting wastewater through the treatment cycle. Traditionally, these stations have relied on fixed-speed pumps, leading to inefficiencies when dealing with variable inflow rates – a common occurrence in WWTPs. Recent advancements in artificial intelligence (AI) offer promising solutions for optimizing pump control strategies and minimizing energy consumption. AI techniques like machine learning and optimization algorithms can analyze real-time data, predict future inflow trends, and dynamically adjust pump speeds accordingly. Studies have demonstrated the effectiveness of AI-based pump control in reducing energy use by significant margins. This research explores a novel AI-based approach combining the Grey Wolf Optimization (GWO) algorithm and Support Vector Regression (SVR) for optimizing pump speed control in WWTP pumping stations. GWO, inspired by grey wolf hunting behavior, identifies optimal pump operation parameters to minimize energy consumption. SVR, a machine learning technique, predicts future pump speed requirements based on real-time data and the optimized GWO parameters. Utilizing real-world data from Zahedan Refinery's pumping station, this study aims to validate the proposed GWO-SVR approach and assess its potential for reducing energy consumption in WWTPs, thereby enhancing operational efficiency and sustainability.In this section, an optimization model for a wastewater pump is developed and presented. The pump is situated in a reservoir where the inflow pattern is randomly generated based on a real-world measured dataset. The Zahedan Wastewater Treatment Plant is located on a 24-hectare land in the east of the city, designed to serve a population of about 900,000 people in the future. The project includes 17 kilometers of transmission lines, a reservoir, and a pumping station. The hourly inflow data to the treatment plant was measured for a pumping station with a constant cross-sectional area of the reservoir of 12 square meters, and a minimum and maximum height of 1 and 9 meters, respectively, in the year 1400 (2021). Based on the 12 month data of 1400, the maximum average wastewater flow rate was 6.99 liters per second in June, and the maximum standard deviation was 68.27 in February (Table 1)The optimization problem is formulated to minimize daily energy consumption for pumping equation(12).Two models are compared:• Classic Model: Fixed pump speed operation• Optimization Model: GWO-based variable speed control for optimized energy usageSVR is introduced to predict pump speed based on data obtained from the optimization model.Discussion and The GWO algorithm is implemented to identify optimal pump operation parameters (Table 3). The SVR model is then employed to predict pump speed based on these optimized parameters. The optimized model is compared to a fixed-speed system, demonstrating significant energy savings. Figures 4 to 6 show the changes between constant-speed (ηcs) and variable-speed (ηopt) performance based on α and β for different time windows of Zahedan station in 2021. With increasing inflow rates (decreasing α), the variable-speed pump performance improves over the constant-speed pump, widening the gap between ηopt and ηcs. Conversely, with decreasing inflow rates (increasing α), this gap narrows. The parameter β, which regulates pump speed, increases performance for both pump types when raised. These optimizations were performed for 30, 60, and 120 minute time windows in 2021, with the variable-speed pump outperforming the constant-speed pump in all cases. Across all time windows, increasing the inflow rate (smaller α) results in a larger energy saving (ε) for the variable-speed pump over the constant-speed pump. With increasing β, which allows higher pump speeds, ε decreases for the same inflow rates (constant α). This is because at higher β values, the variable-speed pump behaves more like a constant-speed pump. The correlation coefficients comparing these optimization results to simulations in 2021 are 0.9996-0.9999 for pump speed and 0.9976-0.9999 for energy, indicating the high accuracy of the SVR simulations and GWO optimization. Figure 10 compares the optimized GWO and simulated SVR results for pump speed in a 60-minute window, while Figure 11 shows the same for pump energy consumption. The close agreement between the optimization and simulation demonstrates the effectiveness of the proposed GWO-SVR model. This study investigated a novel AI-based approach combining the Grey Wolf Optimization (GWO) algorithm and Support Vector Regression (SVR) model for optimizing pump speed control in wastewater treatment plant pumping stations. The GWO-SVR model demonstrated significant energy savings in the variable-flow pumping station compared to a fixed-speed system. The SVR model accurately predicted optimal pump speeds based on real-time data and GWO-optimized parameters, enabling efficient, dynamic pump control. The combined GWO-SVR approach provides a practical solution for WWTP operators to optimize pump control, achieve substantial energy savings, and contribute to more sustainable and cost-effective wastewater treatment practices. Its successful implementation in existing pumping stations can drive improved operational efficiency and environmental sustainability.Keywords: Energy efficiency - simulation - pumping station - gray wolf algorithm - support vector regression

Today, with the unprecedented growth of artificial intelligence (AI) and its increasing applications across various fields, its use in industry has gained significant traction. A notable example of this application is the use of AI techniques for modeling and predicting power requirements in machining processes. The current state has heightened the need for adopting power prediction models to facilitate sustainable process planning. This paper introduces two AI modeling techniques—Artificial Neural Networks (ANN) and Support Vector Regression (SVR)—for predicting power consumption in machining processes. To evaluate the capabilities of these techniques in forecasting power values, a real machining experiment was conducted. The experiments were designed using the Taguchi method to study the impact of all parameters with the minimum number of trials possible. A Taguchi L16 4-level 3-factor design was employed to describe the experimental setup. The predicted power values from both techniques were compared and assessed, revealing that the ANN performs slightly better than SVR. To evaluate the models' performance, several representative hypothesis tests were conducted, including t-tests for mean testing, F-tests, and Levene's test for variance testing. The results indicate that the proposed models in this study are suitable for power prediction.

This paper presents a powerful two-step method for damage detection of large-span bridges with variable sections. Bridges are one of the basic infrastructures in the field of urban and suburban transportation, and timely detection of damage during its operation is important. Damage in this category of structures will cause service disruption during natural disasters. The presented method is based on the combination of spectral finite element and modal strain energy damage index, as well as the combination of genetic algorithm and support vector regression to detect and estimate the damage severity. One of the efficient methods in the field of wave propagation is the spectral finite element method, which is capable of modeling with high flexibility and detecting micro damage. Vibration-based methods are widely used to detect structural damage, while the modal strain energy damage index has a higher sensitivity in detecting damage among other vibration-based methods. The case study model is the Crowchild Bridge in Western Canada, which has special characteristics in terms of geometry and the characteristics of structural elements. In this research, the modal strain energy damage index has been modified due to the change of cross-section along the girders. Also, support vector regression has been used as a robust technique in estimation damage severity. In order to increase the accuracy and improve the damage severity estimation method, the genetic algorithm is used to optimize the effective parameters of the support vector regression. The combined method of genetic algorithm and support vector regression has been able to estimate the severity of damages in a favorable way.

Solar radiation, the basic parameter required for solar energy programs, is not measured in all parts of the world due to technical and financial limitations, and it is necessary to determine with the accurate estimation methods. In this research, a model was developed to estimate the daily horizontal scattered solar radiation with the hybrid methods of Support Vector Regression, Gray Wolf Optimization and Harmony Search Algorithm. To determine the validity of the model, daily measured solar radiation data of different cities in the sunny part of Iran (Bandar Abbas, Kerman, Sanandaj, Semnan and Zahedan) were used. The input parameters of the models are the mean temperature, relative humidity, sunshine hour, evaporation and wind velocity. The comparison of the solar radiation of the hybrid model with the measured values shows the desired results based on statistical analysis, and the hybrid model is an efficient and accurate method compared to other models, especially experimental models. The average absolute bias error obtained, root mean square error and correlation coefficient for Zahedan station are respectively equal to 14.77 MJ⁄m^2, 26.85 MJ⁄m^2, and 0.68 for the experimental equation and the obtained values of the regression vector machine - Harmony respectively It was similar to 13.85MJ⁄m^2, 9.58 MJ⁄m^2 and 0.71. The research results showed that the regression-harmony vector machine model is an efficient method that is much more accurate than other models.

Bridges are exposed to various micro and small damages during their service life. Due to the importance and role of these structures in transportation on roads, timely detection of damages in them has a special place. This research presents an innovative and efficient method for diagnosing and estimating micro and small damages to bridges. The proposed method is based on the combination of spectral finite element and modal strain energy-based damage index to determine the location of damages in the first step. Then the support vector regression technique is used to estimate the severity of damages in the second step. A new eight-node element with spectral finite element characteristics is defined in OpenSees software. Then, micro and small damages are modeled in a single-span beam and a double-span steel beam with spring supports. In the first step, after modal analysis of the structures, the modal strain energy-based damage index is calculated to determine the location of the damages. In the second step, using the modal strain energy-based damage indices calculated in the previous step, the support vector regression networks are trained to estimate the severity of the damages. The results show very high accuracy and optimal performance of the proposed method.

Villages located in Isfahan province are one of the areas prone to the spread of cutaneous leishmaniasis, which is characterized by the occurrence of wounds on the skin. To predict the future prevalence of cutaneous leishmaniasis, Continuous monitoring of the spatial distribution of this disease is essential. Disease modeling was performed using two machine learning algorithms called support vector regression (SVR) and multilayer perceptron neural network (MLP). The performance of these algorithms is evaluated using the RMSE index. Analysis of the results shows that SVR algorithm with RMSE = 0.170 compared to MLP with RMSE = 0.348 has better performance. Environmental factors include temperature, humidity, precipitation, altitude and wind speed as independent variables and Estimation of leishmaniasis density was used as a dependent variable in the modeling process, Of which (70%) were used for model training and the remaining (30%) for model evaluation. The results of spatial analysis index showed that The distribution pattern of cutaneous leishmaniasis in the years 1397 to 1399 was clustered.

The Short-term forecasting of electric load plays an important role in designing and operation of power systems. Due to the nature of the short-term electric load time series (nonlinear, non-constant, and non-seasonal), accurate prediction of the load is very challenging. In this article, a method for short-term daily and hourly load forecasting is proposed. In this method, in the first step, the intrinsic mode functions (IMFs) of the Electric load curve, which are a group of average and pseudo-periodic average signals, are extracted by using the empirical mode decomposition (EMD) method, which is a non-linear and non-constant time-frequency method. For this purpose, the maximum and minimum points of the signal are determined, and then, in one cycle, the difference between the average curve of the upper and lower envelope is calculated with it. This continues until the result falls below a threshold value, and then, the rest of the signal which contains noise is discarded to get a relatively clean signal. In the second step, we need to obtain the sub-sequences of each IMF. So, we use the wavelet transform. The wavelet transform is a kind of transform that is used to decompose a continuous signal into its frequency components, and the resolution of each component is equal to its scale. Each subsequence contains different information and details that can help the improvement of the prediction accuracy. In the third step, the obtained subsequences are aggregated and finally used for prediction by Support Vector Regression (SVR). Support vector regression is a type of supervised learning system that is used for both grouping and estimating the fitting function of data in regression problems so that the least error occurs in the grouping of data or in the fitting function. The purpose of the proposed method is to reduce the error for daily and hourly load prediction. In this method, two datasets of Poland and Canada have been experimented. With four criteria of mean square error (MSE), root mean square error (RMSE), average absolute percentage error (MAPE) and mean absolute error (MAE), the results are evaluated. The findings show that the load prediction error for the Polish data set are as follows: MSE equal to 0.0012, RMSE equal to 0.0342, MAPE equal to 2.9771, and MAE equal to 0.0044. For Canadian data set, the results are as follows: MSE equal to 5.0969e-07, RMSE equal to 7.1393e-04, MAPE criterion equal to 0.9571, and the MAE criterion equal to 2624e-04. Comparison of the proposed method with other competing methods show that better results are achieved by the proposed method in term of the error rate.

Due to the difficulties of conducting tests, especially in weak rocks and the cost of these experiments, by examining the relationships between their mechanical and physical properties can be provided and reduce the cost of tests to identify mechanical properties (Minaeian and Ahangari, 2013; Azadan and Ahangari, 2014).     In this study, petrographic, physical and mechanical experiments on 62 cores of shale and marl of Gurpi Formation were conducted in Godar-Khosh dam site, west of Iran. Non-destructive tests were performed on cores according to the ISRM standard. Physical properties such as water absorption, density and porosity of the samples were determined according to the ISRM standard. Also, uniaxial compressive strength (UCS) test according to the ASTM standard D2938 (ASTM, 1986) was performed. For each sample, the modulus of dynamic elasticity (Ed) and the dynamic Poisson ratio were calculated (Goodman, 1989). Using statistical analysis, artificial neural network (ANN( and  support vector regression (SVR) with radial base kernel function, several relationships for estimating UCS, Es and shear wave velocity were presented. The root mean square error (RMSE), the mean absolute percentage error (MAPE) and the variance account for (VAF) were also used to evaluate the results.

In this paper, a method is presented for predicting time series. Time series prediction is a process which predicted future system values based on information obtained from past and present data points. Time series prediction models are widely used in various fields of engineering, economics, etc. The main purpose of using different models for time series prediction is to make the forecast with the greatest accuracy. The model presented in this paper is based on the combination of kernels and support vector regression. Support vector regression is highly capable of solving function estimation problems by using its kernels, but kernels’ parameters need to be adjusted. First we have preprocessing phase which includes normalizing data and separating data for testing and training. In proposed model, ten different kernels were used. Five kernels were selected as the best kernels by trial and error and these kernels are applied to data. There probably is only a few of the kernels that are useful for the problem, and we are not aware of which kernels are useful for our problem so kernel outputs aggregate by applying a coefficient. This combination creates a new secondary space. The output is given to support vector regression to construct a model that predicts values exactly ɛ accurate, which means the predicted values do not deviate more than ɛ from the original data. This model predicts values by using a leave one out model. Each kernel has parameters that need to be set to optimum values in order to get the best results. Hence in the proposed model, the kernel parameters and their weights are learned by the Gray Wolf Optimizer. This optimizer has been able to provide appropriate answers to many problems, especially challenging problems and has a superior ability to solve the high-dimension problems. By running program in consecutive iterations and examining the different values of the parameters, the optimizer learns the best of them which prediction error has been reduced, and finally returns their best value. The proposed model is implemented on five standard time series and compared to other method, test based on the RMSE criterion for DJ time series, improved by 1.58 point, Radio time series, improved by 0.178 point, and Sunspot time series, improved by 1.709 point. Finally, we analyzed the results, Statistical evaluation by Wilcoxon Signed-Rank Test where the p value is very low compared to the proposed method and CNN-FCM, AR_ model per scale, Multiresolution AR model and ANN methods, slightly lower for Wavelet-HFCM and ANFIS methods and slightly lower than one for SAE-FCM method and at the end provide a relation to find the window size in the model by obtaining the average of peak differences, valley differences, and consecutive peak, and valley differences for the actual values of the training data in exchange for their sequence number in time series.

Load forecasting is one of the most important aspects of the effective management of power systems and depends on various factors such as the means and sources of electricity production by each company, electric load demand, weather factors, economic factors and human activities.. The main goal of this article is the STLF in the regional electricity grid of Guilan province. In this research, after collecting data including wind speed, relative humidity, temperature, dew point, holidays, day length and the effect of corona disease, pre-processing is done on them and they are divided into five clusters using K_Means clustering algorithm. In the following, the relationships between the independent and dependent variables of electricity consumption in Guilan province, using linear regression (LR) algorithms, artificial neural network (ANN) and support vector regression (SVR) along with the Grid search optimization method, were investigated and analyzed in Python software and modeled in the Google Colab environment. The results show that the SVR algorithm has higher accuracy and longer implementation time than the two algorithms of ANN and LR.

The aim of this research is to construct and validate an optical sensor with a multi-beam ratio technology and artificial intelligence-based modelling (MLP & SVR) for suspended sediment measuring. After implementation of the new technology, the performance of the device was evaluated in two stages including calibration and validation. To attain this, various experimental tests carried out in the hydraulic laboratory of Water Research Institute of the Ministry of Energy. Reference turbidity meter and total suspended solids (TSS) were used to test the performance of the OBS technology. In calibration stage, 70% of TSS data were used and the remaining 30% of data were used to validate optical technology. The plotted calibration curves show a very good correlation between the optical voltage recorded by the sensors and the suspended sediment concentration. Also, SVR & MLP models were employed to improv results of suspended load prediction. The performance of optical device and also optical device with intelligence models were evaluated through four statistical indices namely, Mean Absolute Percentage Error (MAPE), Root mean square Error (RMSE), Nash–Sutcliffe coefficient (NSE), correlation coefficient (R) and coefficient of determination (R2). The results of this stage showed that the intelligence modelling could result in improvements of suspended load reported by optical technology. The best improvements obtained by MLP-optical technology. The results showed that values of validation indicators for MLP model are equal to 0.023, 7.608, 0.99, 0.99 and 0.99 respectively, which indicates the proper performance of the technology.

Efficient forecast of precipitation and temperature with a one-month horizon can provide managers with an exceptional opportunity to plan water resources and deal with floods and droughts. The application of proper post-processing and bias correction methods can much improve the accuracy of these predictions. In this study, the S2S (Sub seasonal to Seasonal) precipitation and temperature forecasts of ECMWF were evaluated in one of the important basins of Iran. A variety of methods were used for post-processing and bias correction of these predictions, and the results were compared with different evaluation criteria. Quantile mapping (QM), Bayesian model averaging (BMA), Support vector regression (SVR), an Empirical equation for bias correction of temperature, and some hybrid methods were applied to forecasts. The BMA outperformed the other methods in improving both temperature and precipitation forecasts. Raw precipitation and temperature forecasts were only applicable in 2 or 3 months of the year, but post-processing methods were able to accurately improve precipitation in half of the months, especially rainy months. The hybrid of empirical equation-BMA in 10 months of the year was led to better results than the estimate of the next month's temperature using climatological data.

The prediction of the intrusion of saline water into coastal aquifers as a result of changing the amount of groundwater extractions is a prerequisite for managing groundwater. This study investigates the capability of different types of Support Vector Regression (SVR) models to predict salinity concentrations at the selected well in the small coastal aquifer under different groundwater abstraction conditions. SVR models were trained and tested using input (random transient pumping from the production wells) derived from Latin Hypercube Sampling and output (salinity concentration at the selected well) datasets. The trained and tested models were then used to predict salinity concentrations at the selected well for new pumping datasets. The models ability for predicting and generalizing compared with commonly used artificial neural network (ANN) model was evaluated using different performance criteria. The results of the performance evaluation of the models showed that the predictive capability of the polynomial SVR model is superior to other models. Also, comparing different performance criteria for all SVR models, except for linear SVR model, proved their acceptable predictive performance. The prediction and generalisation ability of polynomial SVR, recommends using these models to connect to the optimization algorithm for a surrogate model based simulation-optimization approach in sustainable management of coastal aquifers.